The Data Scientist Show

Tommy Dang is the Co-founder and CEO of Mage, a data ingestion and transformation pipeline for data engineers (https://github.com/mage-ai/mage-ai). Previously, he was working on data engineering and machine learning engineering at Airbnb. He has a bachelor degree of science in UC Berkeley studying economic, history, and sociology. Today we’ll talk about how he learned engineering and machine learning after college, data tools and ML tools he built at Airbnb, performance review, and how he navigates his career. If you like the show subscribe to the channel and give us a 5-star review. Subscribe to Daliana's newsletter on www.dalianaliu.com/ for more on data science and career.

Tommy’s LinkedIn: https://www.linkedin.com/in/dangtommy/

Daliana's LinkedIn: https://www.linkedin.com/in/dalianaliu/

Daliana's Twitter: https://twitter.com/DalianaLiu

(00:00:00) Introduction 

(00:01:28) Get into computer science from non-tech background 

(00:03:08) How he started his first project 

(00:04:07) Projects at Airbnb 

(00:06:09) Speed vs Quality when building data pipelines 

(00:16:34) How to deal with AdHoc requests 

(00:21:00) How did he learn machine learning 

(00:24:04) How he convinced data scientists to teach him ML 

(00:25:15) Performance review 

(00:27:11) Don’t let your job title limit your career 

(00:28:29) Why he started his company 

(00:31:38) Build your own tool vs use open source solutions 

(00:33:12) Transitioning from an engineer to a CEO 

(00:34:50) Earn trust from internal stakeholders 

(00:36:27) Career advice 

(00:41:31) How he carved his own path at Airbnb 

(00:46:00) How did he learn to be a good engineer 

(00:47:10) Best advice for data scientists or engineers 

(00:48:41) Most important quality of data scientists or engineers 

(00:51:51) Design principles 

(00:58:51) Future of tools 

(01:01:00) What does he think about his future career 

(01:05:05) Inspiration of Tommy

Gabriel Bayomi is the Co-Founder at OpenLayer, a tool that tests & debugs machine learning models. OpenLayer was in the YCombinator’s batch in 2021, building tools for machine learning model testing. Previously he was a machine learning engineer at Apple working on Siri. He has a master degree in computer science from Carnegie Mellon. He is passionate about Natural Language Processing, Machine Learning, and Computational Social Science. We talked about how to test and debug machine learning models, his experience at Apple, and career lessons. If you like the show subscribe to the channel and give us a 5-star review. Subscribe to Daliana's newsletter on www.dalianaliu.com/ for more on data science and career.

Gabriel’s LinkedIn: https://www.linkedin.com/in/gbayomi

Daliana's LinkedIn: https://www.linkedin.com/in/dalianaliu/

Daliana's Twitter: https://twitter.com/DalianaLiu

(0:00) Intro

(01:01:39) How he got into machine learning

(01:06:43) His experience at Apple, Siri

(01:15:55) How to validate the solution

(01:19:39) Benefits of using external error analysis framework

(01:21:30) How to build a model evaluation pipeline

(01:28:26) Don’t overfit the subset of data

(01:33:19) Your validation set shouldn’t be fixed

(01:41:03) Become one with data

(01:44:05) Three model interpretability library you should use

(01:50:47) Common mistakes people made in model validation

(01:53:33) How to create an adversarial test

(01:55:43) How to check data quality

(01:06:46) Transition from engineer to executive

(01:10:04) Things he learnt from his favorite coworker

(01:17:57) how job roles would evolve

Alisa Kim is the head of data product at Vestiaire Collective. Previously, she was a research scientist at Amazon Web Services. We used to work on the same team in Machine Learning Solutions Lab and Amazon Web Services. We have collaborated on projects before and previously she was a consultant and worked on analytics and investment banking. She has a Ph.D. in Econ AI and she has worked on various industries and multiple continents. She's someone I really enjoyed working with. We talked about her journey, the projects she worked on and the lessons she learnt. If you like the show subscribe to the channel and give us a 5 star review. Subscribe to Daliana's newsletter on www.dalianaliu.com/ for more on data science.  

Alisa's LinkedIn: https://de.linkedin.com/in/alisakolesnikova

Daliana's LinkedIn: https://www.linkedin.com/in/dalianaliu/

Daliana's twitter: https://twitter.com/DalianaLiu

(0:00) Intro

(00:01:38) how she got into data science

(00:04:38) day-to-day at AWS ML Solutions Lab

(00:08:00) AWS leadership principles

(00:16:34) challenges the consultant faces when working with external customers

(00:23:36) from AWS to Vestiaire Collective

(00:37:54) how to build a better data product

(00:44:17) how data scientist can align with business stakeholders 

(00:57:52) from tech to business

(01:01:33) how to develop communication skills

(01:09:17) increase visibility of the data science team

(01:17:22) being proactive vs being passive in chasing opportunities

(01:24:06) get feedback from your "nearest neighbors"

(01:25:37) how to set boundary at work

(01:38:48) mistakes she made in her career

(01:48:25) how to manage disagreement

(01:57:53) future of data science

Mark Freeman is a community health advocate turned data scientist His mission is to improve the well-being of people, especially among those marginalized. He is currently a senior data scientist at Humu where he builds data tools that drive behavior change to make work better. He has a master degree from the Stanford School of Medicine in clinical research, experimental design and statistics. He also has a certificate in entrepreneurship from the Business School of Stanford. In his free time, he volunteers with a Bay Area Community Health Advisory Council. He also plays Men's Division III Rugby. We talked about the building data tools, data engineering skills for data scientist, how to pitch a projects, and his career journey. If you like the show subscribe to the channel and give us a 5-star review. Subscribe to Daliana's newsletter on www.dalianaliu.com/ for more on data science.

Daliana's LinkedIn: https://www.linkedin.com/in/dalianaliu/

Daliana's Twitter: https://twitter.com/DalianaLiu

Mark's LinkedIn: https://www.linkedin.com/in/mafreeman2/

Chapters:

(0:00) Intro

(00:03:05) Our experience using R - 1000 lines of code

(00:09:22) Entrepreneurship within a company

(00:16:25) DBT and modern data stack

(00:20:15) Tools don’t matter (in interviews)

(00:21:09) Things DE enjoys but DS doesn’t

(00:24:55) How to work with different stakeholders

(00:30:32) Common SQL mistakes

(00:33:34) SQL vs Python vs R

(00:35:26) T.R.I.B.E framework for projects

(00:40:43) Meet the stakeholders where they at

(00:42:40) Use feedback to get buy-in from collaborator

(00:46:36) How to pitch a new idea

(00:49:45) Don’t lead with solution, lead with the problem

(00:51:03) How to get buy-in from the leadership

(00:57:56) Present an idea as if the audience came up with it

(00:58:41) How to iterate a project

(01:00:27) How he almost lost 1 Million dollar for his company

(01:02:07) Things he learned from his manager

(01:04:19) Things that help people make changes effectively

(01:06:05) Things he learned from mentoring

(01:12:19) Mental Health and anxiety

(01:17:12) Web3

(01:20:14) Why he cares about community health

(01:25:40) "Soul - searching" on his future

(01:28:36) Why he write on LinkedIn

(01:30:04) Future of data science

Abhi Sharma started his career as a software engineer at Amazon Lab 126, building cloud services for Alexa. Later he transferred to Amazon Web Services as a deep learning architect. We used to work at the same team at machine learning solutions lab in AWS. Currently, he is a product manager, responsible for machine learning products like chatbot at Chime. We talked about how he transitioned his career from software engineer to deep learning architect and to a product manager, cool projects he worked on, and our shared experiences at Amazon. If you like the show subscribe to the channel and give us a 5-star review. Subscribe to Daliana's newsletter on www.dalianaliu.com/ for more on data science.

Daliana's LinkedIn: https://www.linkedin.com/in/dalianaliu/

Daliana's Twitter: https://twitter.com/DalianaLiu

Abhi's LinkedIn: https://www.linkedin.com/in/abhivs/

Highlights:

(0:00) Intro

(00:01:48) from SWE to deep learning architect to product manager

(00:12:44) day-to-day as a product manager at Chime

(00:19:46) how he collaborates with different data personas

(00:27:21) how to negotiate for more time for projects with leaders

(00:33:59) some timelines are negotiable

(00:38:00) most impactful project he worked on

(00:44:22) how to evaluate KPI, and not game the system

(00:48:02) think about development in the beginning

(00:50:29) data scientists need to educate the business and demystify the buzz words

(00:54:19) Amazon’s Think Big Challenge

(00:57:09) Never solve the problem twice

(01:00:25) How to transition to a product manager

(01:07:48) why he wanted to become a PM

(01:25:35) How can data scientist learn from PM

Mikiko Bazeley is a senior software engineer working on MLOps at Intuit. Previously, she worked as a growth hacker, data analyst in Finance, then become a data scientist, and later transitioned into machine learning. She has a bachelor degree in econ, biological anthropologie, did data science bootcamp at springboard. She is a tech writer for NVIDIA and she’s working on a course on MLOps. Her goal is to demystify MLOps & show how to develop high-quality ML products from scratch. You can find her content on Linkedin and YouTube. Today, we’ll talk about useful engineering principles for data scientists, MLOps, and her career journey. Subscribe to www.dalianaliu.com for more on data science and career. If you like the show subscribe to the channel and give us a 5-star review. Subscribe to Daliana's newsletter on www.dalianaliu.com/ for more on data science.

Daliana's LinkedIn: https://www.linkedin.com/in/dalianaliu/

Daliana's Twitter: https://twitter.com/DalianaLiu

Mikiko's Linkedin: https://www.linkedin.com/in/mikikobazeley/

Highlights:

(0:00) Intro 

(00:02:00) from GPA2.6 to data scientist

(00:05:27) her experience at Mailchimp

(00:11:44) her frustrations on Cookiecutter project

(00:14:09) the pain point of a data scientist working with engineering

(00:21:01) 2 MLOps pattern

(00:25:52) challenges about her work

(00:29:49) the basic engineering skills a data scientist should have

(00:32:46) the tests a data scientist should write

(00:37:42) how an MLOps engineer collaborates with a data scientist

(00:45:28) what makes a good MLOps engineer

(00:52:33) AWS vs GCP vs Azure

(00:58:59) how a data scientist collaborates with an MLOps engineer 

(01:05:19) suggestions for building a model on a large scale

(01:09:11) how she learnt MLOps on her own within 6 months

(01:17:32) learn from code review

(01:19:17) MLOps books and resources she recommended

(01:24:13) mistakes she made earlier in her career

(01:31:29) common mistakes people make during career change

(01:38:22) "Start with the end in mind"

(01:41:16) the future of MLOps

(01:46:23) how she sees her career growth

(01:56:40) how she continues learning new skills

(02:00:09) what she is excited about her career and life

Max Sklar is an independent engineer and researcher. Previously, he was an engineering and Innovation Labs Advisor at Foursquare after 7 years at the company as a machine learning engineer. Previously, he has worked on Ad Attribution, recommendation engine, ratings. He is the host of The Local Maximum podcast. Max studied CS from Yale, and holds a Master degree in information systems from New York university. If you like the show subscribe to the channel and give us a 5-star review. Subscribe to Daliana's newsletter on www.dalianaliu.com/ for more on data science.

Daliana's LinkedIn: https://www.linkedin.com/in/dalianaliu/

Daliana's Twitter: https://twitter.com/DalianaLiu

Max's Linkedin: https://www.linkedin.com/in/max-sklar-b638464/

Max’s website: localmaxradio.com/about

Interviews he mentioned during the podcast:

Andrew Gelman, Statistics at Columbia University

Shirin Mojarad on Causality

Johnny Nelson on Free Speech and Moderation online

Stephanie Yang talking about Foursquare's Venue Rating System

Dennis Crowley: on Labs, on Innovation

Sophie Carr (Bayesian Mathematician)

Will Kurt (Bayesian)

Marsbot for Airpods

Other Episodes Mentioned

Bayesian Thinking

P-Hacking

Interview on Learn Bayesian Statistics

Highlights:

(0:00) Intro

(00:01:23) from computer science to machine learning

(00:05:35) Bayesian methods in rating system

(00:14:53) how to choose a Bayesian prior

(00:20:10) how to deal with p-hacking

(00:26:57) causality model in ad attribution

(00:35:20) Bias-correction methods

(00:45:43) negative lift in advertising

(00:51:05) unexpected consumer behaviors

(00:52:08) why he decided not to climb the "engineer ladder"

(00:56:46) the challenges of having 5 managers in a year

(01:01:38) using the 3rd-party software vs building his own

(01:04:18) how he approaches ML problems

(01:07:51) his tech stack

(01:09:25) his advise on learning machine learning

(01:12:40) projects he is working on

(01:17:10) Bayesian for his life decisions

(01:22:00) how writing helps him

(01:23:48) the confusion, stress and excitement in his career

Damien Benveniste is a data scientist and software engineer. Previously, he was a machine learning tech leader and mentor. He has worked for almost ten years in different machine learning roles in different industries such as AdTech market research, e-commerce and health care. He has a Ph.D. in physics from Johns Hopkins University and now working towards co-founding own startup in employee engagement space. We talked about his career journey, how he solved challenging problems, and his advice for new data scientists and engineers. If you like the show subscribe to the channel and give us a 5-star review. Subscribe to Daliana's newsletter on www.dalianaliu.com/ for more on data science.

Daliana's LinkedIn: https://www.linkedin.com/in/dalianaliu/

Daliana's Twitter: https://twitter.com/DalianaLiu

Damien's Linkedin: https://www.linkedin.com/in/damienbenveniste/

(00:00) Intro 

(00:01:17) from quantitative trading to machine learning 

(00:07:52) his experience at Meta 

(00:21:16) automated machine learning 

(00:28:52) model paradigm 

(00:32:47) the productivity-oriented culture at Meta 

(00:41:42) short-term gain vs long-term goal 

(00:44:38) things he liked at Meta 

(00:51:54) the project that shaped his career 

(01:03:56) the importance of having a baseline for ML models 

(01:09:12) why he time-boxed everything 

(01:16:25) test the model in production 

(01:20:05)experimental design for ML 

(01:23:25) the most challenging project he worked on 

(01:37:07) best practices for machine learning 

(01:48:44) how he sees himself 

(02:00:52) lessons he learnt from being layoff 

(02:06:45) frustration he had in his previous job 

(02:16:14) what he is working on 

(02:29:18) the future of machine learning 

(02:39:52) things he is excited about

Sunishchal Dev is a lead data scientist at Booster. He's helping to decarbonize the transportation industry by optimizing last mile delivery of renewable fuels. Previously, he was a management consultant. On the side, he volunteers with Project Drawdown to model the most effective solutions to climate change. He is also a mentor of future data scientist as a springboard by guiding them through real world projects. We talked about his career journey, supply chain optimization, how data science can help the environment. If you like the show subscribe to the channel and give us a 5-star review. Subscribe to Daliana's newsletter on www.dalianaliu.com/ for more on data science.

Daliana's LinkedIn: https://www.linkedin.com/in/dalianaliu/

Daliana's Twitter: https://twitter.com/DalianaLiu

(0:00) Intro

(00:01:24) from business to data science

(00:06:36) the big impact of a small improvement

(00:08:50) data engineering vs predictive modeling

(00:11:48) routing optimization

(00:16:27) time series model

(00:21:32) use upsampling to simulate intermittent time series problem

(00:26:20) his modern data stack

(00:28:29) collaborate with engineers

(00:30:06) common mistakes people made in building time series model

(00:37:02) collaborate with truck drivers

(00:40:17) how to become a good communicator

(00:46:30) his experience in mentoring data scientist

(00:51:14) things people cannot learn at school

(00:53:16) the mistakes he made and the things he learnt from his mentor

(00:56:07) how data science can help the environment

Books recommended: 

The Pyramid Principle: Logic in Writing and Thinking

The Book of Why: The New Science of Cause and Effect

Influence, New and Expanded: The Psychology of Persuasion

Felicia Rutberg is a product strategy and analytics manager at Snap, previously she was a product data scientist at Spotify. She started her career as a management consultant at Accenture. She studied mathematics and cognitive psychology at the Vanderbilt University. Felicia reached out to me on Linkedin because she wanted to share how she became a data scientist while having ADHD. Today we’ll talk about product analytics at Spotify and Snap, her career journey, and ADHD. If you like the show subscribe to the channel and give us a 5-star review. Subscribe to Daliana's newsletter on www.dalianaliu.com/ for more on data science.

Daliana's LinkedIn: https://www.linkedin.com/in/dalianaliu/

Daliana's Twitter: https://twitter.com/DalianaLiu

Felicia's Linkedin: https://www.linkedin.com/in/feliciarutberg/ 

Highlights: 

(00:01:29) from management consulting to data science 

(00:12:20) financial data analyst at Spotify 

(00:20:06) how to do internal job transition 

(00:25:57) product data scientist at Spotify in the econometrics team 

(00:29:33) how she became more vocal on the creative process

(00:33:48) how to get the last 1% of the work done 

(00:38:53) how to ensure the quality of the analysis 

(00:50:19) propensity score matching at Spotify 

(00:57:09) how to validate causal inference outcomes 

(01:00:51) lessons from working with economists 

(01:19:16) from Spotify to Snap 

(01:27:35) salary negotiation 

(01:34:02) day-to-day at Snap 

(01:38:33) Spotify vs Snap 

(01:44:35) lessons from management consulting that helped her data science journey 

(01:47:37) ADHD and self-compassion 

(02:02:52) the books she recommended 

(02:08:26) her future career