Skip to content

Comet ML Office Hours 13 | 09MAY2021

May 13th, 2021

48 mins 10 secs

Season 12

Your Host

About this Episode

Comet provides a self-hosted and cloud-based meta machine learning platform allowing data scientists and teams to track, compare, explain and optimize experiments and models.

Backed by thousands of users and multiple Fortune 100 companies, Comet provides insights and data to build better, more accurate AI models while improving productivity, collaboration and visibility across teams.

Register for future sessions here: http://bit.ly/comet-ml-oh

Checkout Comet ML by visiting: https://www.comet.ml/

Checkout the latest FREE e-book from Comet - Building Effective Machine Learning Teams: https://bit.ly/3bWrJ0O

Or on Twitter: https://twitter.com/CometML

On YouTube: https://www.youtube.com/channel/UCmN63HKvfXSCS-UwVwmK8Hw

Vote in the data community content creators awards! http://bit.ly/data-creators-awards

Check it out and don't forget to register for future office hours: http://bit.ly/adsoh

Register for Sunday Sessions here: http://bit.ly/comet-ml-oh

Watch the episode on YouTube here: https://www.youtube.com/playlist?list=PLx-pFw_ty92wJoWzoO7WlfaM7iYB8_qjm

Checkout the interview I had with the Super Data Science Podcast: https://www.superdatascience.com/podcast/landing-your-data-science-dream-job

Transcript

comet-ml-ma9.mp3

Speaker1 00:09

What's up, everybody? How's it going? Welcome. Welcome to

Speaker2 00:12

The comet and our office hours

Speaker1 00:14

Powered by the [email protected] and have all of you guys here it

Speaker2 00:17

Is,

Speaker1 00:18

Sunday, May 9th. Happy Mother's Day to everybody in the audience. I don't think anybody in this room is a mother, but you might have mothers in your life. So I want to wish everybody happy Mother's Day.

Speaker2 00:29

It was

Speaker1 00:29

My son's first birthday

Speaker2 00:31

Yesterday, so that was a fun experience.

Speaker1 00:34

We did a a drive by birthday party because of the restrictions. People can't really come over and party and we can't really host an event. So we just had people

Read Full Transcript Hide Full Transcript

Speaker2 00:42

Drive up and passed out goodie bags and treated those up for, uh, for some gifts, I guess. But it was a good time. So shout out to my son. Happy birthday. Hi, guys.

Speaker1 00:53

Welcome to the officers. How's it going? Got Kevin Martin

Speaker2 00:56

Tours in the room. Guys, how's it going?

Speaker3 00:58

Yeah, it's another day.

Speaker2 00:59

Yeah, right.

Speaker1 01:01

Well, if you guys got any questions at all, go ahead. Let me know.

Speaker2 01:04

Go ahead and go for it. Kevin, how's it going?

Speaker1 01:07

You're still not part of datastream. Did you try to log in there?

Speaker3 01:12

No, I haven't tried logging at all. I was disconnected when I couldn't pay anymore. I've been unemployed since April of twenty twenty, so I have not been able to get a job anywhere on it. So it's been crazy. So yeah, I mean, I took the Google Analytics course. I've taken I mean, I continue to try to learn everything I can.

Speaker1 01:35

So how did you find that Google Analytics course?

Speaker3 01:38

I don't remember. It's all now. I mean, because I follow a lot of people on Facebook and stuff that are in Data sites like you and Kyle, and now it's super Data Science Club and a bunch of stuff. So I'm sure it came up one place only thing I haven't done like my capstone project. Shame, shame, shame.

Speaker1 02:01

So what type of roles are you out there looking for?

Speaker3 02:03

I'd like to be Data manager. And the reason is, is I have enough technical knowledge and understanding of all the technology involved was a DBA for eight years. So I understand database's. I understand the data and modeling and all that kind of stuff. So I understand enough. And then I've been a project manager, scoutmaster for twenty five years. I'm used to building teams. I build up things really well. I feel I'd be a great team leader for Data science. I don't want people that can just lead teams. They want people that can do all hands on stuff and that's not me.

Speaker2 02:38

So yeah, I

Speaker1 02:40

Mean Data management, Data governance, all that stuff is

Speaker2 02:42

Super, super important. Part of the process. That's like the entire Data strategy kind of stuff is very important. And we need people who know

Speaker1 02:52

How to do that

Speaker2 02:53

Because a

Speaker1 02:54

Data scientist can't really do their work without

Speaker2 02:56

Having good data. Right. So that's one thing I've been really working on

Speaker1 03:00

At my current job over the last month or so is is the data management type of stuff. So it

Speaker2 03:05

Begins with just interviewing a bunch of

Speaker1 03:07

People internally, finding out what they

Speaker2 03:08

Want and what they need and trying to deliver on that. But what's the job search process currently like?

Speaker1 03:14

Like what's where are you targeting? How are you targeting jobs where you mostly look at

Speaker3 03:18

My resumes on those monster and LinkedIn and I have been both just applying for jobs posted on LinkedIn or somebody has posted an announcement for job and paying them and talked to them. And I have tried everything and anything. You know, I now have like five resumes to try to make sure that they're targeted to the jobs and stuff like that.

Speaker2 03:45

So what's your

Speaker1 03:47

Search process like on LinkedIn? I do like keyword searches or is it just a. Yeah. So I guess let me let me try to reframe what the question or what the what the issue might be. Is it a like a are you able to find jobs that you feel are interesting

Speaker2 04:02

To you that you want to go for. But they're, they're not converting

Speaker1 04:05

Into like interviews and opportunities

Speaker2 04:07

And things like, oh you know,

Speaker1 04:09

So explain what you're which processes like. So you're messaging people on

Speaker2 04:12

Linkedin, sending

Speaker1 04:13

Out the resume

Speaker3 04:14

And applying for jobs. Do I mean going through the application process? If it's a job I really like and then I'll go ahead and ping or email the person that posted the job to give them a little bit more about me or stuff like that.

Speaker2 04:29

So but and how is the response rate on that zero.

Speaker1 04:35

Yeah. What what do you say in your in your message, have you tried testing different messages

Speaker2 04:39

And and

Speaker3 04:41

I haven't, I haven't done that yet. Yeah.

Speaker1 04:44

So I mean the biggest thing

Speaker2 04:46

I could I could say is

Speaker1 04:47

Definitely continue to reach out to people, but make sure you're targeting the right people on LinkedIn to reach out to the person who posted the job. That's usually a good person to reach out to, but also going into like the LinkedIn AIs the companies kind

Speaker2 04:59

Of LinkedIn page

Speaker1 05:01

And then filtering that down by like

Speaker2 05:03

Recruiter, technical recruiter and even like. Manager or higher up people in the Data food chain, I guess, and just because animal credits are, you know, you can burn through

Speaker1 05:18

Them rather quickly. So you would make sure that whoever it is that you're targeting is actually somebody who's active

Speaker2 05:22

On LinkedIn quite regularly. So when you

Speaker1 05:25

When you go to your I'll pull it up just to show you guys. So if anybody is actually listening to this on the podcast,

Speaker2 05:32

I'm going to YouTube right now. I mean, YouTube and what a

Speaker1 05:35

Linkedin right now. So you go to YouTube to see what it is that I'm doing. But let's just say, for example, here, let's go look at jobs here. Let's let's make this a real case study. Where are you looking for jobs at?

Speaker3 05:45

Charlotte, Washington, D.C., or really only to Florida or Florida? Those three opportunities up

Speaker1 05:53

Put Washington, D.C. in the search just because it's because and you can search by title skill keywords and so forth. Right. So something you might want to do

Speaker2 06:02

Is Data man.

Speaker3 06:04

Yeah, well, I use Data period, but yeah.

Speaker1 06:08

Yeah, I make it, you know, as target as possible. Right. So Data management or even Data governance

Speaker2 06:14

Is another way we could put that here. Right. And here.

Speaker1 06:18

Let's see what what happens here. So what I'm

Speaker2 06:20

Going to show you here is how I would go about applying for a job. Right. OK, so here I

Speaker1 06:26

See enterprise Data policy, governance, senior sociological. Let's say that I am interested in this

Speaker2 06:30

Job like oh by the number in the description. Looks awesome. Cool. Does it have a job poster in this case.

Speaker1 06:37

Does not look like it. So I'll go to the company website in this case, Fannie Mae, which

Speaker2 06:41

I think is a huge excited credit type of place who I owe a lot of student loans to. Um, yeah, no kidding. But you've got two people, right?

Speaker1 06:52

And you search, you can search by title keywords also.

Speaker2 06:55

So I would go technical recruiter, OK, just because those are the type of recruiters I would tend to hire for Data type of roles.

Speaker1 07:03

And then when you go to technical recruiter, you'll see that, you know, they'll probably have a bunch of people pile up, obviously on target, people who are

Speaker2 07:08

In the location where you're applying for and bunch of people come up.

Speaker1 07:14

Right. So how do I know who to target? Right. I've got like five email credits and I want to maximize the probability of my message actually

Speaker2 07:21

Getting returned and maximize

Speaker1 07:23

The probability that that message is going to result in

Speaker2 07:26

A phone call at least. Right.

Speaker1 07:28

So I could go through each person kind of

Speaker2 07:30

One by one, and I'd go directly to the page here and I'd look at their LinkedIn activity. And this is kind of like a lightweight version of stalking on people, but it's all good. But I don't know why my Internet is so slow, but you go to their profile and typically, like one of the first few boxes up here

Speaker1 07:50

Is right under the about section is going to be

Speaker2 07:52

Activity. And if somebody, you know, you could see how often they've posted, how often they've liked or shared or done something on LinkedIn. Right. So in this case, this guy has activity here.

Speaker1 08:03

And we could see when he was last liking and posting stuff. Right. Because if it was within the last day or two, then that's probably a good person.

Speaker2 08:11

They got a warm lead, so to speak, that you can

Speaker1 08:14

Reach out to. Right. So this guy is in about a week ago. OK, well, to me, you know, I'd look for somebody who's more active

Speaker2 08:20

Within a day or so. Right. And so this guy doesn't really look active.

Speaker1 08:24

I wouldn't waste any more credit on him. I just make sure that you're

Speaker2 08:26

Tailoring OK, call

Speaker1 08:28

Resume itself. I don't think I mean enough advice on what to do with a resume. I think it'll all cancel out to zero eventually. But I am lazy, so I just have one

Speaker2 08:38

Resume and I just have one eye for all jobs. But I'll tailor my message that I reach out to with you. No more custom tailored.

Speaker1 08:48

I guess

Speaker2 08:49

There's all kinds

Speaker3 08:50

Of things that'll that'll help with something I haven't done before, so I appreciate it. And now you can. Let's let Martin ator bug, you know.

Speaker1 08:58

Yeah, definitely. I mean, I know a lot of people are getting value from this as well, even the ones listening on the podcast. But just finish my thought on that real quick. There's another website called

Speaker2 09:09

Eyo where you can

Speaker1 09:11

Essentially get the email pattern

Speaker2 09:13

For a given company to to see what their email standard is like and try to reach out to people inside the company directly

Speaker1 09:22

To the email. I wouldn't reach out to

Speaker2 09:24

Like a data scientist or

Speaker1 09:25

A hiring manager. That way I'd reach out to recruiters

Speaker2 09:28

Because that's their job to do that.

Speaker1 09:30

So definitely do that. And then there's a few emails that you can get on. I've got this email list that I'm on that

Speaker2 09:36

Gets you know, I get a bunch of messages every day about openings,

Speaker1 09:41

See if I could find one really just to give you the name of that

Speaker2 09:44

Company. But there's

Speaker1 09:46

Like listservs and stuff that you can get

Speaker2 09:47

On in this case. This company is called AIs Solutions Inc. That's oh,

Speaker3 09:55

Yeah. Yes, there are there are big company. And I don't care for AIs solutions that much because most of the times they're recruiting for another company. OK, they're always just discounting the rates. Because they're third parties.

Speaker2 10:09

Yeah, yeah, yeah, but I

Speaker1 10:11

Mean, so hopefully that's that's helpful, like, just make sure that the message is clear. Like I've got a standard kind of template that I will manipulate.

Speaker2 10:20

But the entire

Speaker1 10:21

Thing I'm doing with my message and I reach out to somebody is just telling them what I could do for them, not just telling them how awesome I am,

Speaker2 10:27

But like talking about how my awesomeness is going to benefit them. So that's a subtle shift.

Speaker1 10:32

But thanks for the question. We'll continue. We want feel

Speaker2 10:35

Free to hop on with any other questions at any other point, if you'd like. So let's go to either I'll put you on on mute here, Kevin. Just minimize background noise there. Yeah.

Speaker1 10:48

Let's go to Martin. Martin is a new face. I don't think I've seen my tore. My good friend here is he's definitely only facings been around.

Speaker2 10:55

Martin, how's it going?

Speaker4 10:57

I'm fine, thank you.

Speaker2 10:58

Are there any questions or anything? Let me know.

Speaker4 11:01

Um, my question is how to to to to keep on learning what what what is the next step for me. So I started to learn Python two or three years ago and then I come to to Data signs. I did a lot with Data and learned did you pendas course. And actually I think it's time to to enter the machine learning thing. But I'm I'm not sure how to start this for.

Speaker2 11:49

Yeah. I'd start by doing like a like

Speaker1 11:51

Many projects as much as possible just to start developing the skill set, developing and understanding of the workflow,

Speaker2 11:58

So and so forth. So I just

Speaker1 12:00

Kind of backtrack here a little bit. So give me a little

Speaker2 12:02

Better sense of your background here. So you so

Speaker1 12:05

You currently you've been studying Python for three years.

Speaker2 12:08

Like, do you have like a Data background, stat's, background, math, background or anything like that?

Speaker4 12:13

Well, I'm actually a civil engineer and I'm working with long term traffic models since 15 years or so. Yeah. So so there are some Data which I use all the time, but not not very in a statistical way, just applying some functions or changing some Data to get to court and prediction. Yeah.

Speaker1 12:47

So that when you say you want to get into machine learning, is it like you've never studied it before, you just kind of wanna get your feet wet with it to learn and understand some of the common algorithms and what

Speaker2 12:56

This thing is all about? Yeah, it's about OK, so I'm a big fan of books. My favorite book is called

Speaker1 13:04

Hands on Machine

Speaker2 13:05

Learning with Python

Speaker1 13:07

Or Secular Hands on Machine Learning. This I could learn and tensorflow.

Speaker2 13:11

That's a good one. And just this

Speaker1 13:13

Introduction to machine learning. Those are good Texel. I'll pull them up in a second here. But once you go through one of those tags, I would say the next thing is try to find the intersection of civil engineering

Speaker2 13:23

And machine learning. So as you were speaking, I pulled up real quickly, just a

Speaker1 13:30

Quick Google search, civil engineering, machine learning. And it looks like there's, you know, machine learning techniques for civil engineering.

Speaker2 13:35

A primer on machine learning applications in civil engineering.

Speaker1 13:38

So you want to study machine learning and civil engineering, right. So I would try to find the intersection between

Speaker2 13:44

Machine learning, civil engineering, because you've got extensive experience in civil engineering and now you're trying to blend that with

Speaker1 13:51

Machine learning. That just makes for easier

Speaker2 13:54

Learning experience, in my opinion, more enriched learning experience.

Speaker1 13:58

So I can give you a link to to this. Most of these type of

Speaker2 14:01

Papers will kind of walk you through

Speaker1 14:03

A general review process of what machine learning is all about.

Speaker2 14:06

Um, so I'll link to this as well. But the two books I was talking about in particular there was introduction to machine learning with Python. Right.

Speaker1 14:17

And it's this one right. Here is the O'Reilly

Speaker2 14:19

Text, the O'Reilly text here. But if you

Speaker1 14:23

Prefer online courses. So, yeah, I guess such as what

Speaker2 14:26

Type of matter you like to consume your

Speaker1 14:29

Your educational content. Are you more of a book scholar.

Speaker2 14:31

More of a.

Speaker4 14:33

Yes, I like, I like online courses but often they are not going deep enough for me. Yeah I did some, I started one machine learning course but it was only shown how it is applied in Python. So just but but not what, what is going on behind all of this. OK.

Speaker1 15:00

Oh yeah. OK, so those two books will, they'll give you mostly just a familiarity with the secular and a. But they probably will go in depth in the math

Speaker2 15:10

As it seems like you want to go. So for that I would there's a free book actually that that is available, I think it's called and I'll get

Speaker1 15:19

You the link for it if you go to my LinkedIn. PDF must have shared it recently, but it was like

Speaker2 15:23

Immediately there's Bitly. I think it was like mathematics or something like that. What did I make the. That might not be it,

Speaker1 15:33

But I can I can give you the

Speaker2 15:34

Link

Speaker1 15:36

In a second here. I'll remember

Speaker2 15:37

What it is, but it's a

Speaker1 15:39

Book that goes really, really

Speaker2 15:40

In depth. So, you know, let me just try to find it real quickly.

Speaker1 15:45

But that's what that's what my approach would be. I'll pause here to see if there's any questions from you or any comments while I

Speaker2 15:51

Try to start to dig this link up for you. But but please go for it. Martin, if

Speaker1 15:56

You got any follow up questions

Speaker2 15:57

Or comments, let me let me know.

Speaker4 15:59

Um, yeah, I heard some from Kiru, you know, I guess. Yeah. And I heard he said that it is better to to start projects than to learn just by courses.

Speaker2 16:16

Yeah.

Speaker1 16:16

Projects are definitely, definitely the way to go.

Speaker2 16:18

Right. So um so once

Speaker1 16:20

You like the foundation knowledge you definitely

Speaker2 16:23

Start doing actual projects.

Speaker1 16:27

But the projects I would recommend you do are ones

Speaker2 16:29

That that are rooted with civil

Speaker1 16:31

Engineering. So there's that intersection of civil engineering and machine learning just because it make it so much more enjoyable. You might even find opportunities for you

Speaker2 16:38

To, uh, to employ it in your current job. Other projects are 100 percent like the best way to learn. Um, but I mean, you have to at some point start from a bit of source material. Right.

Speaker1 16:50

So I found the link here. It's mathematics for machine learning.

Speaker2 16:53

It goes super, super in depth. Um, so just a warning. Let me get you link here. All right. I'll pull it

Speaker1 17:01

Up real quickly as well. Sorry for everybody listening on the

Speaker2 17:04

Podcast, stalling here for a second. But yeah, this

Speaker1 17:07

Book is called Mathematics for Machine Learning, and it goes super

Speaker2 17:09

In-depth with all the

Speaker1 17:11

Behind the scenes math and stuff, if that's

Speaker2 17:13

What you're interested in. I know some people like to, uh, look,

Speaker1 17:17

It's not satisfying enough just to see what what the API does. It is good to see what's

Speaker2 17:21

Going on behind the

Speaker1 17:22

Scenes of this book is really, really good for that. So when Models meets Data

Speaker2 17:26

And a few other things and I can give you a few other links, but I absolutely like by all means, do projects, you must must do projects to, uh, to

Speaker1 17:35

Really solidify understanding. And when you do projects,

Speaker2 17:38

You post them on like GitHub. Right. And you can share them as proof of work, put them on your resume when you start looking for jobs. So these are all added benefits of of doing projects. Let me know if that was

Speaker1 17:51

Helpful or not or if I was just being too vague.

Speaker2 17:54

I'm happy to be to.

Speaker1 17:57

And then let me give you a

Speaker2 17:58

Link to this paper that's doing machine

Speaker1 18:01

Learning techniques for civil

Speaker2 18:02

Engineering problems. Um, I think that would be really helpful. Again, just because

Speaker1 18:07

You're already so knowledgeable about civil engineering that seeing how this new thing applies in your

Speaker2 18:12

Domain would make that learning curve a little bit easier. And then here's

Speaker1 18:17

Another book as well, Machine Learning

Speaker2 18:18

In Action, and I'll give you that link here. Awesome. Cool.

Speaker1 18:23

Well, Martin, let me know if there's any other questions as we progress.

Speaker2 18:26

I'm happy to, uh, to take them. Let's continue. Go got there a. in the House and two years ago when,

Speaker1 18:33

Um, Granada is also here. Good to see you, Renata. Any questions from anyone?

Speaker4 18:37

Jill, let's find out a vehicle that thinks that.

Speaker1 18:42

Awesome. Well, let's open for prayers for whoever else would like to go for it.

Speaker2 18:46

So, Jill, Renata, um, a.. No questions from Jill. Awesome. To have to go and get any questions or comments.

Speaker3 18:54

No questions. Just a small comment. I was just looking at that book you just listed. I got to page six, seven, eight, all Greek. To me,

Speaker1 19:06

It's literally literally all Greek. Yeah. That that first link, that mathematics from machine learning, that is definitely for for people who want

Speaker2 19:13

To know what the hell is going on under the hood, which

Speaker1 19:17

Sounded like Martin wanted that type of reference. But the other one I gave him I think is a little, uh,

Speaker2 19:22

More lightweight, not lightweight.

Speaker3 19:24

You have to be into math and really understand formulas, I guess.

Speaker2 19:28

Yeah, definitely. Definitely an important part of Data science machine learning for sure

Speaker3 19:33

By itself isn't really covering up to that level. So I would like to go. Yeah, that's going to be my biggest challenge. You know, it's one thing I can't mention. I understand in principle some that I'm starting to get now, but to go from that and then actually being able to utilize those skills, that means I would have to kind of go back on my path all over again in a few years.

Speaker2 19:56

Yeah, I mean, but that ground up approach to like it might

Speaker1 20:00

Work for some people, might not work for some people, but

Speaker3 20:02

It won't work for me. I'm a hands on. So yeah. Yeah. Get my people up and running. That's the first step. Yeah, you know, trial and error till you kind of get the basics. And that's how I approach things like,

Speaker2 20:15

You know, I like that approach.

Speaker1 20:17

Hands on is, you know, you always learn better by doing on the

Speaker2 20:20

Job type of type of thing, so. Exactly.

Speaker1 20:24

Let's see if we had any questions that we had a..

Speaker2 20:27

Or not A, no questions from a..

Speaker1 20:31

Just been grinding, studying on evenings and weekends

Speaker2 20:33

And gradually also

Speaker1 20:35

Using Data since I was

Speaker2 20:36

A little stream project at the moment. Yeah. So as much as you can try to

Speaker1 20:40

Find opportunities to use Data science

Speaker2 20:42

Methodology in your current role, I highly, highly recommend that when I was trying to learn Python.

Speaker1 20:47

It's like four years ago. So at that time I was a biostatistician. I was working

Speaker2 20:51

As a essentially clinical trial statistician at a pharmaceutical company. All of our work was done in SACE, S.A.S., because that's what we would have to

Speaker1 21:01

Submit code along with the Data

Speaker2 21:03

To the federal government, to the FDA. And I wanted

Speaker1 21:07

To learn how to use Python. So what did I do? I recreated everything that I didn't

Speaker2 21:12

Sace in Python

Speaker1 21:13

Because I had like a basis of

Speaker2 21:14

Comparison. Right. Like I knew that my Sasko did this and

Speaker1 21:19

Manipulated Data this way and had

Speaker2 21:20

This particular output. Cool. Now let me see if I can do that in Python. And I did that

Speaker1 21:25

And just gradually started

Speaker2 21:26

Just building the skills that way. And it was it was a bit of a learning curve. Took me two, three months to

Speaker1 21:34

Get decent at it.

Speaker2 21:35

I'm still not like I'm

Speaker1 21:36

Not a software engineer by any

Speaker2 21:37

Means, but I can get stuff done. I get a lot of stuff done. But that's one good way to learn, is just

Speaker1 21:42

Find opportunities in your current role to leverage and use this stuff. All right.

Speaker2 21:46

Sacia Renata or anybody else, go questions. Go for it. Jill is

Speaker1 21:51

Heading out. All right. So what else is going on, guys? We got we got some we have some time here, man.

Speaker2 21:56

I'm happy to happy to chat about

Speaker1 21:58

Anything not even related to Data seems to be

Speaker2 22:00

Completely honest.

Speaker1 22:01

Denseness is cool and all, but it's not like the only thing I like talking about.

Speaker2 22:05

So if he has any questions on on anything at all,

Speaker3 22:09

You want to help me. You want to help me write. My dissertation

Speaker2 22:13

Was a dissertation about

Speaker3 22:15

Social mobile social media marketing for small businesses in Charlotte.

Speaker2 22:19

That's pretty interesting. Tell me more about that mobile

Speaker1 22:22

Social media marketing.

Speaker3 22:24

Yeah, you know, first of all, Web advertising has only been around since ninety four, so that's generally pretty early thing. And mobiles only really been around since around twenty seventeen. The mobile searches on Google have increased over desktop searches, so for years mobiles have been kind of king on search. But the problem is that most small businesses lose. That's my opinion. I'll find out when I actually did the research is they don't know how to do it. They're not equipped to do it or trained to do it. And even though they may want to do it for the business. So I found somebody who all he did was put this business on Google Maps, happens to be in Texas, but it's still a thing. And according to him, their income increased threefold by only going on to Google Maps only for small business.

So that's the type stuff is to see, you know, if a small business takes the steps to go ahead and go on Google Maps or advertise on Facebook or Instagram, I know there's millions of places they can advertise trying to limit it so I can get my dissertation done, you know, what's out there, what can they do and try to increase their sales or their customer base or, you know, one is just how are you? Like, give an example. We went out to a restaurant and it was the food was the and the service stunk. And I put a nasty Yelp comment on for that restaurant because that restaurant bothered to even contact me or respond. Nope. So they've got this really bad review sitting out there and they haven't bothered to respond, which I don't think is proper. I think they should have at least responded to it anyway.

Speaker1 24:15

Yeah, yeah. Mobile mobile marketing is definitely interesting. Like marketing in general is just super interesting. I to be fascinated not just the not just like the advertising

Speaker2 24:24

Part of it, like doing the

Speaker1 24:26

Analytics on the ads and stuff like that. That's interesting. But just marketing

Speaker2 24:29

It, you know, as a field I find really, really interesting. A huge fan of Seth Godin.

Speaker1 24:35

He's written a few books on marketing. This is marketing.

Speaker2 24:37

All marketers are liars, but he's he's awesome. Yeah. Read Ryan Holiday's book called Perennial Seller,

Speaker1 24:47

Which is all about marketing as well.

Speaker2 24:48

So pretty interesting stuff.

Speaker3 24:50

So anyway, that's what I'm doing. That's taking up a lot of my time, as well as trying to look for a job, trying to write dissertations.

Speaker2 24:58

Yeah, man, it's it's hard to look for a job, man. I mean, I could I could identify with that struggle. I mean, I don't

Speaker1 25:05

Know, man. I feel like some people. You like people in interviews,

Speaker2 25:09

They don't ask good questions, right, like I've been on, you know, I've been on tons of

Speaker1 25:14

Interviews, right. Like I've been on tons of

Speaker2 25:15

Interviews and people just

Speaker1 25:19

The interviewers just asking stupid questions that somebody asked me how many models

Speaker2 25:23

I've built or

Speaker1 25:24

How many models have I trained. And I didn't know if it's a trick question, because if you include models trained and

Speaker2 25:28

Cross-validation probably millions. But I mean, like, what's the point of asking

Speaker1 25:34

Me that question?

Speaker2 25:35

Like, that's just really not a very useful

Speaker1 25:39

Question, but also questions like like I had somebody ask,

Speaker2 25:43

Like, you know, fairly recently the interview question was something along the lines of

Speaker1 25:48

This asking me a lot of different things, a lot different people. What type of data scientist are you? And it was a vague

Speaker2 25:53

Question looking for a specific answer, because this position was it is not

Speaker1 25:59

A position where you'd be deploying machine models into production.

Speaker2 26:02

You'd be doing a lot of statistics and analysis and that type of respect. So what he's

Speaker1 26:09

Trying to discern is, are you the type of data scientist that just wants to deploy models into

Speaker2 26:13

Production or do you prefer doing statistical and and probabilistic inference and stuff like that, which he did not mention

Speaker1 26:20

About in the interview, that this what the job was, was going to be until way later?

Speaker2 26:24

They just kind of clicked to me like, oh, OK.

Speaker1 26:26

Well, you're just asking a really weird question.

Speaker2 26:28

Why not just clearly state at the

Speaker1 26:30

Forefront of the interview that this is what the job will entail? Is this the type of work that you enjoy doing as a data scientist?

Speaker2 26:36

Right. So vague questions like that are really, really stupid. I give a shit ton of bad questions, Martin. Go for it.

Speaker4 26:45

How important do you think is it to have a knowledge about field and in that you are doing data science in, you know, can do and in civil engineering are

Speaker2 27:01

Very, very, very important? I think that's that's really, really important. Understanding domain knowledge gives you more

Speaker1 27:07

Context, makes you more knowledgeable and understanding what the implications of what this model is that you're building what it

Speaker2 27:14

Does. Right. If you can understand

Speaker1 27:15

The context to which

Speaker2 27:16

It will be used. So domain knowledge is important. Is it like a barrier to getting a job? Probably not, unless you're going

Speaker1 27:23

To super

Speaker2 27:24

Specialized, right. If you're going for, like, I don't know, computer vision or or self-driving

Speaker1 27:30

Cars type of machine learning stuff, then, you know, you better really, really understand that domain because I think it'd be hard to break into that.

Speaker2 27:36

It's like the first job. But if you're doing a product analytics and and stuff like that, it's, you know, domain knowledge.

Speaker4 27:43

So I live in an area where a lot of companies are located and they are looking for Data science very often. But I don't think that is my field.

Speaker1 27:58

Yeah, so they're probably rebranding biostatisticians as Data scientist or something like that. I'm not sure, but I feel like that you'd have to learn a lot of

Speaker2 28:06

Regulatory rules, especially pharmaceuticals. When I was in pharmaceuticals, there's so much like

Speaker1 28:12

Regulatory stuff that I

Speaker2 28:13

Needed to to know, mostly with respect to the work that I'd be authoring like protocol sections, statistical

Speaker1 28:20

Sections of protocols or statistical analysis

Speaker2 28:23

Plans and things like that. Um, it's helpful. Um, I mean, like, why not just apply for and see what happens is there's never any harm with that.

Speaker4 28:31

Do you know do you know any data science project in the field of mobility laboratory.

Speaker2 28:40

A little bit more. What you mean by mobility.

Speaker1 28:42

Um, I just like people moving around the mobility. I would have mobility.

Speaker4 28:47

No, not. And new mobility like like how to predict traffic or OK to give people information about if there is any trouble ahead when they are driving in the car or public public transport being.

Speaker2 29:14

Yes. I mean

Speaker1 29:15

The first thing I do is just start researching

Speaker2 29:17

The, um, the domain.

Speaker1 29:20

Right, because there's I mean, I get

Speaker2 29:22

A lot of knowledge from just reading white

Speaker1 29:25

Papers and research papers because you go to the reference section and there's some great references that that kind of help you,

Speaker2 29:30

You know, build a chain of good resources, but just start

Speaker1 29:35

Googling. That's the easiest thing. Right. So we're talking about here like,

Speaker2 29:37

I know, traffic prediction. Right. So we do traffic prediction and machine learning.

Speaker1 29:43

And let's just see if somebody has actually done this in a

Speaker2 29:46

Jupiter notebook and

Speaker1 29:48

See what comes up. Right. So in this case, I'm doing traffic prediction machine learning

Speaker2 29:52

Jupiter notebook and we've got traffic sign detection for Tensorflow is probably not what you're talking about. Um, yeah. So let's remove this, uh, file type part or even look for white papers. Um, so you

Speaker1 30:04

Had to start looking for research papers and see what happens. So we got deep learning on traffic prediction, I call comparison of machine learning methods

Speaker2 30:11

For the air traffic prediction architecture based on machine learning, traffic flow, prediction of big Data machine learning approaches for traffic flow forecasting. All right, cool. So there you go. And it looks like there's a ton

Speaker1 30:23

Of research in that area. I would just dig into it by getting one of

Speaker2 30:27

The papers

Speaker1 30:28

That seems most approachable, kind of read through

Speaker2 30:30

It, understand it, and then, you know, dig in a little bit deeper. So, for example, this one that we're looking at, a PDF just opened up because I accidentally downloaded a PDF, I guess I don't know where it went. Here it is.

Speaker1 30:44

And it's a comparison of machine learning methods for the prediction of traffic speeds in urban places.

Speaker2 30:51

So let me just go ahead and pull this up here.

Speaker1 30:53

And most white papers will

Speaker2 30:54

Have, um, I don't know where the, uh, where my screen went. All right. Sorry about this. So I'm pulling up the PDF here just to show you how to go through it. So there are a bunch of people here, right. So I would I would if I

Speaker1 31:10

Were any time I'm trying to learn how to figure out how to do something new, I go to a research paper, I take note of all these people, and I'll add them

Speaker2 31:16

On LinkedIn just to see if

Speaker1 31:18

They're active on LinkedIn or see if they have, like university profile, because maybe I want to connect with them and ask them

Speaker2 31:22

Questions. And, you know, like, hey, your paper is interesting. Here's what I have a question. Right. That's completely OK. But most white papers have the reference section, which are probably the best section. And I mean, it's pretty, really thorough. It looks like this is a really shows you the architecture and everything, which is really cool. Um, but what we're

Speaker1 31:42

Looking for is the references section as well. Right.

Speaker2 31:44

Because you might find more interesting papers that are more specific to what

Speaker1 31:50

It is that you

Speaker2 31:51

Are looking for.

Speaker1 31:52

Another thing you could do is let's say

Speaker2 31:54

You've you've pulled up like two or three or four white papers in the domain.

Speaker1 32:00

Just print out the reference sheet from all these or the references from all these white papers and see what's common among those. And that's a good indication that,

Speaker2 32:07

Ok, that's probably a source of truth that I should look into. OK, but that's that would be my my approach to that.

Speaker1 32:14

So question here from a.. You ask who gets on the podcast, whether they see Data science where they decide to go to the next two to five years. What's my take on it?

Speaker2 32:23

Flipping, flipping the script on me here. So where do

Speaker1 32:26

I think Data says is going in the next two to five years? Definitely don't think it's going

Speaker2 32:30

Away, that's for sure. It's here to stay.

Speaker1 32:33

And I think the next three to five years I don't see anything like

Speaker2 32:37

Really, really earth shattering happening. We might make progress on some interesting things like self-driving cars and things like that. But let me just stick to to the

Speaker1 32:48

Profession of Data science as a self.

Speaker2 32:50

I just see it becoming, you know, more and more discrete kind of roles. Right. I feel like it'll move more towards specialization. Opera operationalization as well. So then talks about this a lot. And I agree with his perspective on this. You'll see more about Ops II Ops type of roles, pop up more architecture type of roles, pop up, that is for sure.

Speaker1 33:13

But I think even beyond five years,

Speaker2 33:15

I think every job is going to

Speaker1 33:17

Become a little bit of

Speaker2 33:18

Data science at some level for

Speaker1 33:21

Many professions, especially professions that do a lot of

Speaker2 33:24

Computation, that do a lot of damn replication work with a lot of Data you'll see

Speaker1 33:31

Them doing more and more Data type of stuff, more and more

Speaker2 33:33

Analytics. I mean, my come to mind, there might be a day where, you know, accountants

Speaker1 33:39

Start using

Speaker2 33:39

Python scripts in their in their Excel formats, in their Excel files, which is possible. Right. Because you can you can integrate Python into Excel. So I see that trend happening more. I think

Speaker1 33:52

People are just there just to be more and more data scientist at work doing data

Speaker2 33:56

Science type of work, but maybe don't have the data science type title, if that makes sense to go for it.

Speaker3 34:00

And I a good comment that you have. I don't disagree with you on that. And I also think that I think the biggest challenge right now for Sassan streamlining Data science into more jobs is the tools. I mean, Python and ah and all these tools. It's not something you just get into and do you have to sit down and then you have to use a regular basis. Now, Excel has had extreme success because it's a tool that everyone uses. Starting off with two plus two equals four and that kind of build gradually. Personally, I think that when they integrate Python etc, I think it will be an indirect arrangement where it basically will still kind of use the Excel. But you will then have more standards that are going to be running scripts for you like macros, etc.. And over time, yeah, doing better answers right now, preparing for an audit. I'm going to be starting at one that week and I'm looking now at 800. Thousand transaction lines. The original Data received that sixty three columns. Now I'm up to about one hundred and forty because I have to add all the extra columns to extract, streamline, adjust the Data.

Now, that part, the part that the title could have done something about or are could have done for me, but that's I can't program it myself that it's just going to have to wait. So I'm still doing it manually. But now Data science is definitely here. We're living on Data Data. That's going to be an hour every day for the rest of our lives.

Speaker2 35:45

Yeah. Yeah. It's not like,

Speaker1 35:47

You know, there's going to be any decrease or slow down to the amount of data being generated around the world. But we got more and more

Speaker2 35:53

Devices that collect data that, you know, companies are hopefully using ethically. That's something

Speaker1 35:59

I'd like to see happen in the next two to five years,

Speaker2 36:02

Maybe more courses on ethics, in data science, how to how to be a data scientist who doesn't cause harm or just some awareness and ethics, um, I think would be very, very useful. But I hope that answered your question that a.. I'm happy to elaborate on that. But to towards point, you know, you

Speaker1 36:23

Just your people are going to have to start

Speaker2 36:26

Doing things involving

Speaker1 36:28

Coding because it's going to make their job easier. Right. They might come across something at work which are like, oh, my God, I've spent so much time doing this.

Speaker2 36:34

I wonder if I can write something, a script or something that can make my life a little bit easier.

Speaker3 36:41

That's that's the low code. No code development system development that I don't know if you've heard about it, but first time I heard about it has been around for like five to seven years or 15, but I've just heard about it where it allows people with no programing background to do that exact same thing. I see something that's too difficult. I want to make it easy so they can go on to Lokodo called platform and they can develop the application they need to do the job. But it's really cool.

Speaker2 37:11

Yeah, yeah. A lot of

Speaker1 37:12

Low code, no code platform. The thing that's great, it makes it easier for people to do stuff. But once you start getting into the edge, cases like these companies are back. Backdoor way out is that if you get to weird edge cases, eventually you have to

Speaker2 37:22

Go into the hood and code and it's usually python code that you're going to have to to rate yourself. So that's kind of like the other way around it.

Speaker1 37:29

Like you still, even if you're using low code, no code environment, like it's not

Speaker2 37:32

Always taking your Desai's jobs, if they're do low code Data science, it's just a different type of person working on different types of problems. But eventually

Speaker1 37:42

Even that

Speaker2 37:43

Person will have to learn how to code because something they're trying to do with their low code, no code environment is bugging out, not working. Um, they have to do write code. And he says this happens to him constantly,

Speaker1 37:56

Since learning are for just over a year, sees things that he used to spend so much time doing.

Speaker2 38:01

Now that is so much faster. Yeah, I mean, even like try to find ways to make your life easier somehow. Right. Like, I had an epiphany earlier this week and it was

Speaker1 38:11

The stupidest epiphany. I've been using computers for a very,

Speaker2 38:13

Very long time. Right. Like I've

Speaker1 38:15

Been using computers since I was,

Speaker2 38:17

I don't know, like twelve, thirteen, maybe thirty eight soon. So I've been using them for quite some time, but I never bothered to learn like shortcuts

Speaker1 38:26

For just the operating system

Speaker2 38:28

And stuff like that.

Speaker1 38:30

And I just recently made the transition to using

Speaker2 38:32

A Mac just like two years ago and never really properly learned the shortcuts for for the Mac OS. And I was spending

Speaker1 38:39

Time like, here's my setup.

Speaker2 38:40

Right? I've got I've got a keyboard and I got a trackpad right. Then it's my laptop right here on the on a stand. But I'll be typing and

Speaker1 38:49

I need to do something. I've spent a lot of time moving

Speaker2 38:51

My hand over to the trackpad moving stuff around. It will cause wrist pain and it'll just take time. And I was like, dude, this is so stupid. Why don't I just learn how to use the shortcuts on the keyboard so then I don't have to keep on going to the track pad.

Speaker1 39:06

And so now I only really use the track pad or try to only use the track

Speaker2 39:09

Pad for stuff that I cannot do with the keyboard, which tends, I think to be more like creative type of stuff. It's the way I'm thinking of it in my head because I like boring type of stuff that I can imagine. A computer being able to write a script for that stuff probably has keyboard shortcuts so bad. That's just just give me a hand.

Speaker3 39:30

So I got another question for you. Yeah, I've heard of descriptive analytics, predictive, prescriptive, and recently I read an article about collaborator's analytics. Are you familiar with it? You know what that entails? I haven't gotten into it in detail. I was just curious if you were aware of that.

Speaker2 39:47

I've never heard that term. But I try to infer maybe

Speaker1 39:51

What they mean by collaborative analytics because there's platforms that like Google Collaboratory

Speaker2 39:57

Or Data breaks where data scientists can work on the same notebook at the same time collaboratively. So if that's. What is meant by that and then I think that's that's awesome, but let

Speaker1 40:09

Me look this up.

Speaker2 40:10

Let me Google this for a clip real quick, collaborative and a little. So you have never heard of this thing before?

Speaker3 40:20

Yeah, it just I mean, within the last couple of weeks, I've read about it.

Speaker2 40:23

So I thought essentially it says

Speaker1 40:26

It combines business intelligence

Speaker2 40:28

Software with some type of

Speaker1 40:31

Increased analytics teams, knowledge. So, yeah, it just looks like it makes it easier to collaborate with maybe non

Speaker2 40:39

Data scientists or non super technical type of people on a project, which I think is great. We need more of that. I agree. We need to learn it right now, probably not just to learn it

Speaker1 40:49

Kind of on the job we need

Speaker2 40:50

To on demand. That's my theory for pretty much everything is just

Speaker1 40:54

Mastered the basics, master the fundamentals and learn

Speaker2 40:57

New things on demand when you need to. Because if you

Speaker1 41:00

Have a solid foundation, solid fundamentals, then you can learn anything you need to. I think I spend a lot of my time now, like reading books that are related to the field, but not textbooks.

Speaker2 41:10

So for example, this book right here that I'm going through the book of why I buy Judea Pearl, it's it's a book. It's a statistics book. But it's not like, you know, a textbook that you go through with, like an online course, a class like that. But these are the books that I feel give me a better intuition of the dry facts that I learned in school, because in school, after grad school, I learned a shit ton of formulas and I learned how to do things with formulas. But I never really understood the history of some of the stuff that were behind there. So these books are really interesting for that purpose. That's why I'm

Speaker1 41:45

Interviewing the author for this one

Speaker2 41:47

Sheep. Jordan Ellenberg also wrote How Not to Be Wrong Power of Mathematical Thinking, New York Times bestseller. I haven't read through this yet. Probably start that later this week.

Speaker1 41:57

But let's go to Joshua.

Speaker2 41:57

So, Joshua, you got a question or anything? Let me know that he was unmetered very briefly or Renato for not as a question. Joshua, go for it.

Speaker1 42:05

See your unmuted, but I'm not able to hear you.

Speaker2 42:08

All right,

Speaker1 42:09

Renata, how are things going with you know, you got a lot of advice and information last week. I'm wondering if

Speaker2 42:16

If you made any progress on those projects you're working on or trying to work on. All right. Does not seem like it looks like Renata's audio is failing,

Speaker1 42:26

But, yeah, let me know if anybody else has questions. Got just a few more minutes here. So I'm happy

Speaker2 42:30

To to stick around. I just want to

Speaker3 42:31

Quickly follow up on my mentioning about, you know, learning command. Similarly, I mean, that's technically what I do. And every day when you get requests, there's probably kind science as well as I always evaluate whether that is something that's going to be a one time thing or is that something that's going to be a regular type, whether it's a report or a summary or some sort of delivery? And given if it is something that I expect that's going to be coming again and again, then I will spend a lot of extra time to properly develop so I don't have to do it again in the future. So I assume it's a similar thing with that science. It could probably do some very quick things, get it resolved. But on the other hand, you can probably if you believe that this is something that's going to be used again, it will probably spend more time to prepare and properly designed something so that it could be used to get out

Speaker2 43:29

Like that approach.

Speaker1 43:30

That's an excellent, excellent approach. And it's all about the master, the basics, master fundamentals like I like.

Speaker2 43:36

And I hear a lot

Speaker1 43:37

From students that they has to him to others like very well. I studied this and I studied this. I studied machine learning. I've studied this algorithm, that algorithm.

Speaker2 43:46

Which algorithm should I learn next? And it's like, how about

Speaker1 43:49

Stop studying algorithms and just try to use one of them

Speaker2 43:52

For a change and compare them and see how that works? I think that's a much more fruitful effort. So looks like Josh

Speaker1 44:00

Is having dashboards, having network issues

Speaker2 44:03

Related to having my

Speaker1 44:04

Problems. But Antia, a question here. Do you have one great book for learning

Speaker2 44:07

Statistics, basics for a non mathematician? I've been reading the comics you mentioned, but looking for a good book to read next. Yeah, I love the comic series, the Cartoon Guides. Those are some of my favorite books, but a good book for

Speaker1 44:24

Statistics, basically for a non

Speaker2 44:26

Mathematician. Naked Statistics by by Charles Wheelan. That's a great book, but I'm not sure if you want more or less like a textbook type of a book. Um, because, I mean, realistically, every textbook type of book will have a bunch of math in it. So it by design.

Speaker3 44:42

Mathematician Statistics for Dummies is an excellent starter statistics book, which is very high level, very easy to understand. And then you can go from there into deeper articles later.

Speaker2 44:54

Yeah, that's a great point.

Speaker1 44:55

Yeah. Those for Dummies books are awesome.

Speaker2 44:56

I love them. I've got a few of them sitting here. I got a few of those books there.

Speaker1 45:02

They break stuff down and make it easy to to understand. So definitely. That's a good recommendation if you're looking for more of a book that's

Speaker2 45:09

Just kind of a nonfiction book about statistics and how it's used

Speaker1 45:14

Widely used

Speaker2 45:15

In a bunch of different fields and domains, then I would recommend

Speaker1 45:19

That. Naked Statistics by Charles Wheelan. That's a good

Speaker2 45:22

One. Yeah, that that's I got a bunch of other ones here. But how not to be right now, how not to be wrong,

Speaker1 45:29

How to measure anything is actually a

Speaker2 45:31

Really good book by Douglas Hubbard.

Speaker1 45:34

A lot of good stats in that book not really designed for mathematicians. More just for for this thinking business people so highly recommend that one as well. But if you're looking for more of a

Speaker2 45:42

Textbook type of thing, then yeah, definitely go for that for Dummies book. Right on. So hopefully everybody is doing well today, which has a very happy

Speaker1 45:52

Mother's Day to all the mothers listening

Speaker2 45:55

To all of you guys, um, the mothers in your lives hanging out with, uh, with my wife and their baby very soon, right after this. Hopefully has had a chance to check out the episode I released from my podcast just a couple of days ago with my friend Prachi Tucker, and that those are the interview I released.

Speaker1 46:16

The officers will be released a little bit later. I didn't get a chance

Speaker2 46:19

To work on them yesterday due to my son's first birthday. So edit and upload those up a little bit later today.

Speaker1 46:26

So if you

Speaker2 46:26

Guys are waiting for that, they'll be available. Got some great interviews

Speaker1 46:32

Coming up for you guys on the podcast in the very near future. Next week is actually with Dennis Rodman. I believe

Speaker2 46:39

He is.

Speaker1 46:41

Interesting guy at the AIs will enjoy that conversation. I really enjoyed talking to him.

Speaker2 46:45

He's an author for Pact, so he's written a bunch of books for Packed, the

Speaker1 46:50

Most recent of which

Speaker2 46:51

Is all about Transformers, which which my friend Tom Ive's is is all about as well. But yeah, that will be released this Friday. Got a bunch of other interesting stuff coming up on the podcast, just to name a few episodes dropping in the coming weeks. We've got an episode releasing with Steve Nory in a few weeks. I've got one with Kingi coming up in a few weeks here. Um, an interview with Chase Caprio. Chase Caprio works for Impact Theory.

Speaker1 47:21

So that's Tom Baillieu's company. And I'm a huge

Speaker2 47:24

Tom Billings fan, so that's a good one. I really enjoyed speaking with him.

Speaker1 47:28

Interviewed Dave Gray, who wrote the book Liminal Thinking, and a bunch of other books. So that was really fun.

Speaker2 47:34

Fun interview as well. And then an interview with the philosopher Jamie Woodhouse. We're talking about this philosophy he developed called sentient ism, which should be

Speaker1 47:43

Interesting, does not look like there are any other questions,

Speaker2 47:46

Like I stalled for as long as possible. So, guys, thank you so much for spending part of your Sunday

Speaker1 47:52

Morning with me. You enjoyed having you here. If there's any last minute questions, now's the time.

Speaker2 47:57

Kevin, any last minute questions? No.

Speaker3 47:59

To say thanks,

Speaker2 48:00

Roland Martin

Speaker1 48:01

Or Tory or anybody else, I don't mind. We got to take care of the rest of the weekend. Remember, you've got

Speaker2 48:07

One life on this planet. Why not try to do some big cheers, everyone.

Download Transcript

Episode Comments

Redirecting to The Employable Data Scientist...