The MLOps Podcast: Recent Episodes
DAGsHub
A podcast about bringing machine learning into the real world. Each episode features a conversation with top data science and machine learning practitioners, who'll share their thoughts, best practices, and tips for promoting machine learning to production
In this episode, I chat with Ljubomir Buturovic, VP of ML and Informatics at Inflammatix. We discuss using ML to diagnose infections and blood tests in the emergency room. We dive into the challenges of building diagnostic (classification) and prognostic (predictive) modes, with takeaways related to building datasets for production use cases.Join our Discord community: https://discord.gg/tEYvqxwhah ---Timestamps:00:00 What is Inflammatix and how do they use ML7:32 Edge Device Deployment: The Future of Model Deployment21:16 Navigating Regulatory Submission for Medical Products 26:01 Evolution of Regulatory Processes in ML for Medical Applications30:18 Challenges and Solutions in ML for Medical Applications34:00 The Future of AI in Clinical Care40:25 The Overrated Concept of Interpretability in AI and ML45:32 RecommendationsLinks🌎📈 Our world in data: https://ourworldindata.org/ 🚀 Profiles of the future: https://www.amazon.com/Profiles-Future-Arthur-C-Clarke-ebook/dp/B00BY7GITK➡️ Ljubomir Buturovic on LinkedIn – https://www.linkedin.com/in/ljubomir-buturovic-798156/➡️ Ljubomir Buturovic on Twitter – https://x.com/ljbuturovic🌐 Check Out Our Website! https://dagshub.com Social Links: ➡️ LinkedIn: https://www.linkedin.com/company/dagshub ➡️ Twitter: https://twitter.com/TheRealDAGsHub ➡️ Dean Pleban: https://twitter.com/DeanPlbn
In this episode, Idan Gazit, Senior Director of Research at GitHub Next, discusses his role in exploring strategic technologies and incubating long bet projects. He explains how the GitHub Next team chooses research projects and the process of exploration and theme selection. Idan also shares insights into the ML focus at GitHub Next and the challenges of evaluating the impact of AI products. He reflects on his journey into the AI space and provides advice for testing AI products in smaller organizations. Finally, he shares his thoughts on the future of AI interfaces.Join our Discord community: https://discord.gg/tEYvqxwhah ---Timestamps:00:00 Introduction and Background00:56 Choosing Research Projects at GitHub Next06:09 ML Focus in GitHub Next10:52 ML Work and the Leaky Abstraction13:16 Idan's Journey into the AI Space17:54 Evaluating the Impact of AI Products24:36 Testing AI Products in Smaller Organizations32:52 The Future of AI Interfaces40:01 Transitioning from Prototype to Product46:45 Challenges in the ML/AI Space56:03 Recommendations➡️ Idan Gazit on LinkedIn – https://www.linkedin.com/in/idangazit/➡️ Idan Gazit on Twitter – https://twitter.com/idangazit🌐 Check Out Our Website! https://dagshub.com Social Links: ➡️ LinkedIn: https://www.linkedin.com/company/dagshub ➡️ Twitter: https://twitter.com/TheRealDAGsHub ➡️ Dean Pleban: https://twitter.com/DeanPlbn
In this episode, I chatted with Uri Goren, founder and CEO of Argmax, about Machine Learning and the future of digital advertising in a world moving away from cookies due to privacy laws like GDPR and CCPA. We chat about challenges in maintaining personalized ads while respecting user privacy, and new methods like probabilistic models and contextual features to cover some of the gap left by removing cookies.Join our Discord community: https://discord.gg/tEYvqxwhah ---Timestamps:00:00 Introduction00:35 The Rise of Privacy Regulations1:40 The Impact of Losing Cookies 2:48 Understanding Cookies4:33 Reasons for the Decline of Cookies8:47 ML Leveraging Cookies in Advertising10:32 The Shift to Contextual Features12:53 The Future of ML without Cookies15:23 New and Old Ways of Generating Contextual Features20:33 Regulatory Conspiracies22:33 Unsolved Problems in ML and AI24:39 Predictions for the Next Year in AI and ML26:17 Controversial Take: Overuse of LLMs28:03 Recommendations➡️ Uri Goren on LinkedIn – https://www.linkedin.com/in/ugoren/🌐 Check Out Our Website! https://dagshub.com Social Links: ➡️ LinkedIn: https://www.linkedin.com/company/dagshub ➡️ Twitter: https://twitter.com/TheRealDAGsHub ➡️ Dean Pleban: https://twitter.com/DeanPlbn
In this episode, I speak with Han-Chung Lee, a machine learning engineer with a lot of interesting takes on ML and AI. We dive into the buzz around natural language processing and the big waves in generative AI. They chat about how newcomers are racing through NLP’s history, mixing old school and new tech, and the shift towards smarter databases. Han-Chung breaks it down with his straightforward takes, making complex AI trends feel like coffee chat topics. It’s a perfect listen for anyone keen on where AI’s headed, minus the jargon.Join our Discord community: https://discord.gg/tEYvqxwhah ---Timestamps:00:00 Intro0:41 State of NLP and LLMs1:33 Repeating the past in NLP3:29 Vector databases vs. classical databases8:49 Choosing the right LLM for an application12:13 Advantages and disadvantages of LLMs16:10 Where LLMs are most useful21:13 The dark side of LLMs and can we detect it?25:19 Thoughts on LLM leaderboard metrics31:19 Using LLMs in regulated industries36:40 Creating a moat in the LLM world40:20 Evaluating LLMs44:20 Impact of LLM on non-english languages48:35 Thoughts on MLOps and getting ML into production56:48 The Hardest Unsolved Problem in ML and AI59:09 Predictions for the Future of ML and AI1:03:25 Recommendations and Conclusion➡️ Han Lee on Twitter – https://twitter.com/HanchungLee➡️ Han Lee on LinkedIn – https://www.linkedin.com/in/hanchunglee/🌐 Check Out Our Website! https://dagshub.com Social Links: ➡️ LinkedIn: https://www.linkedin.com/company/dagshub ➡️ Twitter: https://twitter.com/TheRealDAGsHub ➡️ Dean Pleban: https://twitter.com/DeanPlbn
In this episode, I had the pleasure of speaking with Mila Orlovsky, a pioneer in medical AI. We delve into practical applications, overcoming data challenges, and the intricacies of developing AI tools that meet regulatory standards. Mila discusses her experiences with predictive analytics in patient care, offering tips on navigating the complexities of AI implementation in medical environments. This episode is packed with actionable advice and forward-thinking strategies, making it essential listening for professionals looking to impact healthcare through AI.Join our Discord community: https://discord.gg/tEYvqxwhah ---Timestamps:00:00 Introduction and Background4:03 Early Days of Machine Learning in Medicine 5:19 Challenges in Building Medical AI Systems6:54 Differences Between Medical ML and Other ML Domains15:36 Unique Challenges of Medical Data in ML 24:01 Counterintuitive Learnings on the Business Side28:07 Impact and Value of ML Models in Medicine29:41 The Role of Doctors in the Age of AI38:44 Explainability in Medical ML44:31 The FDA and Compliance in Medical ML48:56 Feedback and Iteration in Medical ML 52:25 Predictions for the Future of ML and AI53:59 Controversial Predictions in the Field of ML56:02 Recommendations57:58 Conclusion➡️ Mila Orlovsky on LinkedIn – https://www.linkedin.com/in/milaorlovsky/🩺MeDS – Medical Data Science Israel Community – https://www.facebook.com/groups/452832939966464/🌐 Check Out Our Website! https://dagshub.com Social Links: ➡️ LinkedIn: https://www.linkedin.com/company/dagshub ➡️ Twitter: https://twitter.com/TheRealDAGsHub ➡️ Dean Pleban: https://twitter.com/DeanPlbn
In this episode, I had the pleasure of speaking with Jason Liu, an applied AI consultant and the creator of Instructor – an open-source tool for extracting structured data from LLM outputs. We chat about LLM applications, their challenges, and how to overcome them. We also dive into Instructor, making LLMs interact with existing systems and a bunch of other cool things. Join our Discord community: https://discord.gg/tEYvqxwhah ➡️ Jason Liu on Twitter – https://twitter.com/jxnlco🤖 Instructor Blog – https://jxnl.github.io/instructor/🌐 Check Out Our Website! https://dagshub.com Social Links: ➡️ LinkedIn: https://www.linkedin.com/company/dagshub ➡️ Twitter: https://twitter.com/TheRealDAGsHub ➡️ Dean Pleban: https://twitter.com/DeanPlbn
Timestamps:00:00 Introduction02:18 Excitement about Machine Learning and AI03:28 Using LLMs as Backend Developers04:22 Building Applications with LLMs07:07 Building Instructor09:30 Thinking in Logic and Design10:33 Validating Data and Building Systems with Instructor11:49 Thoughts About Product and UX in LLMs17:51 Future of Instructor20:25 Misconceptions and Unsolved Problems in LLMs24:57 Improving LLM Applications26:14 RAG as Recommendation Systems29:32 Fine-tuning Embedding Models32:32 Beyond Vector Similarity in RAG39:32 Predictions for the Next Year in AI and ML45:26 Measuring Impact on Business Outcomes47:06 The Continuous Cycle of Machine Learning48:38 Unlocking Economic Value through Structured Data Extraction50:52 Questioning the Status Quo and Making an Impact
In this live episode, I'm speaking with Jinen Setpal, ML Engineer at DagsHub about actually building, deploying, and monitoring large language model applications.
We discuss DPT, a chatbot project that is live in production on the DagsHub Discord server and helps answer support questions and the process and challenges involved in building it. We dive into evaluation methods, ways to reduce hallucinations and much more.
We also answer the audience's great questions.
Join now to take part in our first live MLOps Podcast episode.
I'll be chatting with Jinen Setpal, ML Engineer at DagsHub about his work building LLM applications and getting LLMs into production.
Sign up for the event at the link here:
https://www.linkedin.com/events/7098968036782596096/comments/
In this episode, I had the pleasure of speaking with Yuval Gabay, MLOps Engineer at WSC Sports. Yuval builds better infrastructure and automation for developing, training, and deploying machine learning models at scale, with a focus on video data. We talk about MLOps methodologies, standardizing deployment in the organization, and closing the loop back from production into training.
Watch the video: https://youtu.be/3m__nRuifsQ
Join our Discord community: https://discord.gg/tEYvqxwhah
➡️ Yuval Gabbay on LinkedIn – https://www.linkedin.com/in/yuval-gabay-68963253/ ➡️ WSC Sports – https://wsc-sports.com/
🌐 Check Out Our Website! https://dagshub.com
Social Links: ➡️ LinkedIn: https://www.linkedin.com/company/dagshub ➡️ Twitter: https://twitter.com/TheRealDAGsHub ➡️ Dean Pleban: https://twitter.com/DeanPlbn
In this episode, I had the pleasure of speaking with Hamel Husain. Hamel is a machine learning and MLOps extraordinaire, he was one of the core maintainers of Fast.ai and has worked on ML and MLOps in places like Data Robot, Airbnb, and GitHub. We talk about Large Language Models, the future role of data scientists in the world of LLMs, and Hamel's approach to solving MLOps problems.
Watch the video: https://www.youtube.com/watch?v=3oElMXPkaVs
Relevant Links: 🐦 Hamel's Twitter – https://twitter.com/HamelHusain 🟦 Hamel's Linkedin – https://www.linkedin.com/in/hamelhusain/ ✍️ Hamel's amazing blog: https://hamel.dev/blog/posts/nbdev/
🌐 Check Out Our Website! https://dagshub.com
Social Links: ➡️ LinkedIn: https://www.linkedin.com/company/dagshub ➡️ Twitter: https://twitter.com/TheRealDAGsHub ➡️ Dean Pleban: https://twitter.com/DeanPlbn
In this episode, I had the pleasure of speaking with Almog Baku, a serial entrepreneur, consultant in Cloud, AI Infrastructure and Foundational models. We talk about Kubernetes, Large Language Models (LLMs), how to get them into production, and how data is becoming a more central piece of the ML landscape. We also Discuss Almog's newest project, Raptor ML, which helps ML teams productionize ML pipelines.
Watch the video: https://www.youtube.com/watch?v=DCApRXhXD_w&feature=youtu.beJoin our Discord community: https://discord.gg/tEYvqxwhah
Relevant Links: 🦅 Check out Raptor for ML productionization – https://github.com/raptor-ml/raptor👫 Join the Open AI & Gen AI TLV meetup group – https://www.meetup.com/openai-genai-tlv/📄 Read a very cool LLM paper – https://react-lm.github.io/➡️ Almog Baku on LinkedIn – https://www.linkedin.com/in/almogbaku/➡️ Almog Baku on Twitter – https://twitter.com/almogbakuRecommendation Links:Watch "Foundation" – https://tv.apple.com/us/show/foundation/umc.cmc.5983fipzqbicvrve6jdfep4x3🌐 Check Out Our Website! https://dagshub.com Social Links: ➡️ LinkedIn: https://www.linkedin.com/company/dagshub ➡️ Twitter: https://twitter.com/TheRealDAGsHub ➡️ Dean Pleban: https://twitter.com/DeanPlbn
In this episode, I had the pleasure of speaking with Sherya Shankar, Ph.D. student at Berkeley RISELab. We chat about auto data validation and MLOps. Sherya shares her insights on several interesting topics, including the challenges of automating the data validation process and how to overcome them. We also discuss what makes organizations able to iterate faster in machine learning, and some predictions about the future of machine learning and MLOps.
Watch the video: https://youtu.be/_hi6--H2HugJoin our Discord community: https://discord.gg/tEYvqxwhah Relevant Links: 📃 Moving Fast with Broken Data – https://arxiv.org/abs/2303.06094v1🤖 Operationalizing machine learning https://arxiv.org/pdf/2209.09125.pdf 📒 Operationalizing notebooks https://smacke.net/papers/nbslicer.pdf ➡️ Shreya Shankar on LinkedIn – https://www.linkedin.com/in/shrshnk/ ➡️ Shreya Shankar on Twitter – https://twitter.com/sh_reyaRecommendation Links:📺 The Glory – A Korean Revenge Drama – https://www.netflix.com/title/81519223⛷️ Go Skiing!🌐 Check Out Our Website! https://dagshub.com Social Links: ➡️ LinkedIn: https://www.linkedin.com/company/dagshub ➡️ Twitter: https://twitter.com/TheRealDAGsHub ➡️ Dean Pleban: https://twitter.com/DeanPlbn
In this episode, Dean speaks with Noa Weiss, the wonderful AI & ML consultant. They dive into Deep Learning research for marine mammal sounds, abstractions for machine learning projects and some of the unspoken challenges she's seen in the ML development process. Also prediction markets and Harry Potter.
Watch the video: https://www.youtube.com/watch?v=uQrR0KPq3RQ
Join our Discord community: https://discord.gg/tEYvqxwhah
Relevant Links:
Noa's talks:
🎥 The Quick & Dirty AI Startup: https://www.youtube.com/watch?v=HZ_LRxP3ep0
🎥 Choosing the Right Machine Learning Abstraction for your Business Needs – https://www.youtube.com/watch?v=C0gN47H91HM
➡️ Noa Weiss on LinkedIn – https://www.linkedin.com/in/noa-weiss/
➡️ Noa Weiss on Twitter – https://twitter.com/NWeiss
🌐 Noa's website: https://www.weissnoa.com
Recommendation Links:
-
Unsong – https://unsongbook.com/
-
Harry Potter & The Methods of Rationality – https://www.hpmor.com/
🌐 Check Out Our Website! https://dagshub.com
Social Links:
➡️ LinkedIn: https://www.linkedin.com/company/dagshub
➡️ Twitter: https://twitter.com/TheRealDAGsHub
➡️ Dean Pleban: https://twitter.com/DeanPlbn
In this episode, I speak with Assaf Pinhasi, ML engineering and MLOps consultant extraordinaire! Assaf was the VP R&D at Zebra Medical Vision, and built the PayPal Risk organization's Big Data Platform. We dive into building ML infrastructure from scratch 10 years ago vs. today, best practices involved in building teams to support machine learning models in production, and the future of generative models.
Watch the video: https://youtu.be/tSbuDA5tMxQ
Join our Discord community: https://discord.gg/tEYvqxwhah
➡️ Assaf Pinhasi on LinkedIn – https://www.linkedin.com/in/assafpinhasi/
🌐 Check Out Our Website! https://dagshub.com
Social Links:
➡️ LinkedIn: https://www.linkedin.com/company/dagshub
➡️ Twitter: https://twitter.com/TheRealDAGsHub
➡️ Dean Pleban: https://twitter.com/DeanPlbn
In this episode, I speak with David Marx, Distinguished Engineer at Stability AI. This talk dives into how David got into machine learning, open-source software, and Stability AI. We discuss following your curiosity, and what it takes to deploy a model like Stable Diffusion to production.
Watch the video: https://youtu.be/49dsoDK1KCA
Join our Discord community: https://discord.gg/tEYvqxwhah
Relevant Links:
🛌 Big Sleep by Ryan Murdock – https://colab.research.google.com/drive/1NCceX2mbiKOSlAd_o7IU7nA9UskKN5WR?usp=sharing (Author – https://sigmoid.social/@Adverb)
➡️ David Marx on LinkedIn – https://www.linkedin.com/in/david-marx-b0a5bb14/
➡️ David Marx on Twitter – https://twitter.com/DigThatData
Recommendation Links:
📚 Guerrilla Analytics – https://guerrilla-analytics.net/
💿 Contribute to open source!
🧱 Build lego! It's awesome
🌐 Check Out Our Website! https://dagshub.com
Social Links:
➡️ LinkedIn: https://www.linkedin.com/company/dagshub
➡️ Twitter: https://twitter.com/TheRealDAGsHub
➡️ Dean Pleban: https://twitter.com/DeanPlbn
In this episode, I speak with Logan Kilpatrick, Julia Language Developer Community Advocate. We talk about machine learning at NASA and how he discovered Julia as a student, the age-old Julia vs. Python debate, and how to get into a new scientific and technical field. It was absolutely awesome! Check it out.
Watch the video: https://www.youtube.com/watch?v=3kgRN8hJIro
Join our Discord community: https://discord.gg/tEYvqxwhah
Relevant Links:
➡️ Logan Kilpatrick on LinkedIn – https://www.linkedin.com/in/logankilpatrick/
➡️ Logan Kilpatrick on Twitter – https://twitter.com/OfficialLoganK
➡️ Julia Language on Twitter – https://twitter.com/JuliaLanguage
Recommendation Links:
📚 Three-Body Problem Series: https://www.amazon.com/Three-Body-Problem-Cixin-Liu/dp/0765382032
📹 13 Lives on IMDB: https://www.imdb.com/title/tt12262116/
🌐 Check Out Our Website! https://dagshub.com
Social Links:
➡️ LinkedIn: https://www.linkedin.com/company/dagshub
➡️ Twitter: https://twitter.com/TheRealDAGsHub
➡️ Dean Pleban: https://twitter.com/DeanPlbn
In this episode, I speak with Guy Smoilovsky, my friend, Co-Founder, and the CTO of DagsHub. We talk about quantum computing and AGI, concrete approaches for automating ML deployment, and how DagsHub came to be.
Watch the video: https://www.youtube.com/watch?v=67dByhXPT5g
Join our Discord community: https://discord.gg/tEYvqxwhah
Relevant Links:
➡️ Guy Smoilovsky on LinkedIn – https://www.linkedin.com/in/guy-smoilovsky/
➡️ Guy Smoilovsky on Twitter – https://twitter.com/Guy_T_Sky/
TDD in machine learning – https://towardsdatascience.com/tdd-datascience-689c98492fcc
Recommendation Links:
Astral Codex Ten – https://astralcodexten.substack.com/
Don't Worry About the Vase – https://thezvi.wordpress.com/
The Sandman – https://www.imdb.com/title/tt1751634/
Lady Silver – https://www.ladysilverband.com/
🌐 Check Out Our Website! https://dagshub.com
Social Links:
➡️ LinkedIn: https://www.linkedin.com/company/dagshub
➡️ Twitter: https://twitter.com/TheRealDAGsHub
➡️ Dean Pleban: https://twitter.com/DeanPlbn
In this episode, I speak with Dean Langsam, Data Scientist at SentinelOne and one of the organizers of PyData in Israel. We chat about imposter syndrome, the best field in machine learning, why XGBoost is the best model, and the fact that most organizations have too much data. It was fascinating for me, so I hope you enjoy it too.
🎬 Watch the video: https://www.youtube.com/watch?v=Akz_PpDdLlQ
Join our Discord community: https://discord.gg/tEYvqxwhah
Relevant Links:
⭐️ Join PyData Tel Aviv: https://pydata.org/telaviv2022/ ⭐️
➡️ Dean Langsam on LinkedIn – https://www.linkedin.com/in/deanla/
➡️ Dean Langsam on Twitter – https://twitter.com/dean_la
Recommendation Links:
🌐 Check Out Our Website! https://dagshub.com
Social Links:
➡️ LinkedIn: https://www.linkedin.com/company/dagshub
➡️ Twitter: https://twitter.com/TheRealDAGsHub
➡️ Dean Pleban: https://twitter.com/DeanPlbn
In this episode, I had the pleasure of speaking with Jacopo Tagliabue, Director of AI at Coveo. We talk about Reasonable Scale MLOps, how to approach building your ML platform, and how quickly you might hit the limits of model deployment (hint: it's pretty surprising)
Join our Discord community: https://discord.gg/tEYvqxwhah
Relevant Links:
➡️ Jacopo on LinkedIn – https://www.linkedin.com/in/jacopotagliabue/
➡️ Jacopo on Twitter – https://twitter.com/jacopotagliabue
Recommendation Links:
📺The Boys – https://www.imdb.com/title/tt1190634/
📚Gödel, Escher, Bach – https://www.goodreads.com/book/show/24113.G_del_Escher_Bach
📚The Three-Body Problem – https://www.goodreads.com/book/show/20518872-the-three-body-problem
🌐 Check Out Our Website! https://dagshub.com
Social Links:
➡️ LinkedIn: https://www.linkedin.com/company/dagshub
➡️ Twitter: https://twitter.com/TheRealDAGsHub
➡️ Dean Pleban: https://twitter.com/DeanPlbn
In this episode, I had the pleasure of speaking with Goku Mohandas, founder of Made With ML. Goku has an incredible amount of experience building and teaching the community about machine learning and MLOps systems. We dive system thinking and solving for ML workflow, his journey in the machine learning world, and how he chooses what to learn next. We discuss the most common mistakes he's seen in productionizing ML models and why building models no one will use is not necessarily bad.
Join our Discord community: https://discord.gg/tEYvqxwhah
Relevant Links:
🤩 Check out Made With ML and thank us later – https://madewithml.com/
➡️ Goku on LinkedIn – https://www.linkedin.com/in/goku/
➡️ Goku on Twitter – https://twitter.com/gokumohandas
🌐 Check Out Our Website! https://dagshub.com
Social Links:
➡️ LinkedIn: https://www.linkedin.com/company/dagshub
➡️ Twitter: https://twitter.com/TheRealDAGsHub
➡️ Dean Pleban: https://twitter.com/DeanPlbn
In this episode, I had the pleasure of speaking with Kyle Gallatin, a Machine Learning Software Engineer at Etsy. We talk about how he built the machine learning platform at Etsy, experimentation in production (yes, you heard right), and how to optimize model performance at very large scales. It was awesome, and I'm sure many of you can learn a ton from this one!
Join our Discord community: https://discord.gg/tEYvqxwhah
Relevant Links:➡️ Kyle on LinkedIn – https://www.linkedin.com/in/kylegallatin/
🌐Check Out Our Website! https://dagshub.com
Social Links:LinkedIn: https://www.linkedin.com/company/dagshub
Twitter: https://twitter.com/TheRealDAGsHub
Dean Pleban: https://twitter.com/DeanPlbn
In this episode, I'm speaking with Charlene Chambliss, Software Engineer at Aquarium. Charlene has vast experience getting NLP models to production. We dive into the intricacies of these models and how they differ from other ML subfields, the challenges in productionizing them, and how to get excited about data quality issues.
Join our Discord community: https://discord.gg/tEYvqxwhah
Relevant Links:
- ➡️Charlene on LinkedIn – https://www.linkedin.com/in/charlenechambliss/
- ➡️Charlene on Twitter – https://twitter.com/blissfulchar
Recommendations:
- 🎬3blue1brown – Awesome YouTube channel about math & science: https://www.youtube.com/c/3blue1brown
- 🎙NLP Highlights – Allen AI Insititute podcast about NLP research: https://soundcloud.com/nlp-highlights
- 🎙Software engineering daily: https://softwareengineeringdaily.com/
- 🎙TWiML – Another great podcast about machine learning and AI: https://twimlai.com/
- 📰Sebastian Ruder's blog and newsletter about NLP and ML: https://ruder.io/
- 📰Taming the Tail: Adventures in Improving AI Economics: https://a16z.com/2020/08/12/taming-the-tail-adventures-in-improving-ai-economics/
- 📰State of AI report (2021): https://www.stateof.ai/
- 📕Learn to learn – Ultralearning by Scott Young: https://www.scotthyoung.com/
🌐Check Out Our Website! https://dagshub.com
Social Links:
- 🟦LinkedIn: https://www.linkedin.com/company/dagshub
- 🐦Twitter: https://twitter.com/TheRealDAGsHub
- 🐦Dean Pleban: https://twitter.com/DeanPlbn
In this episode, I'm speaking with the one and only, Yannic Kilcher! We talk about sunglasses 😎, the value and methodologies behind taking complex machine learning research, and making the idea accessible and digestible. We also discuss reproducibility in machine learning and the moving between research and entrepreneurship.
If you haven't seen his videos you should definitely check them out on his YouTube channel (https://www.youtube.com/c/YannicKilcher).
Join our Discord community: https://discord.gg/tEYvqxwhah
Relevant Links:
- ➡️Yannics's amazing YouTube channel – https://www.youtube.com/c/YannicKilcher
- ➡️Yannic on LinkedIn – https://www.linkedin.com/in/ykilcher/
- ➡️Yannic on Twitter – https://twitter.com/ykilcher
Recommendations:
- 🎬Veritasium – YouTube channel about science with really good explanations about complex topics: https://www.youtube.com/c/veritasium
- 📖The Fifth Season – Good sci-fi fantasy book: https://www.amazon.com/Fifth-Season-Broken-Earth/dp/0316229296
🌐Check Out Our Website! https://dagshub.com Social
Links:
- ➡️LinkedIn: https://www.linkedin.com/company/dagshub
- ➡️Twitter: https://twitter.com/TheRealDAGsHub
- ➡️Dean PlbnTwitter: https://twitter.com/DeanPlbn
In this episode, we dive into the challenging but very important topic of getting data scientists to write better code. How to approach complex machine learning projects and break them down, and why growing unicorns 🦄 is better than hunting them. Check out this is an awesome conversation with Laszlo Sragner, Founder at 🔥 Hypergolic.
Join our Discord community: https://discord.gg/tEYvqxwhah
Timestamps:
- 00:00 Podcast intro
- 01:00 Guest introduction
- 02:34 Why is writing better code important for data scientists?
- 03:40 How to improve your code
- 08:17 Don't be afraid of your code.
- 10:42 Breaking experiments into manageable pieces
- 12:35 How did your past experiences teach you to strive for better code?
- 15:21 Proving better code is worth it
- 18:07 What could be adopted from software development
- 23:06 What's the most interesting/challenging part of taking models to production?
- 27:12 What is the hardest part about building a machine learning model?
- 29:30 How it looks when it works well – a detailed example
- 36:23 The difference in writing better code in smaller startups compared to larger organizations
- 39:18 Laszlo's process for the first iteration in a machine learning project
- 44:33 Breaking data problems down into vertical slices
- 47:55 End-To-End Platforms vs. Best-of-breed tools
- 50:30 Obligatory job title discussion...
- 53:30 Hunting for data science unicorns
- 56:33 Traits to look for when building a data science team
- 58:30 Build vs. Buy? What's better?
- 59:56 What is the most exciting trend in ML and MLOps?
- 1:00:47 How do you stay up to date?
- 1:01:40 Recommendations for the audience
Relevant Links:
- ➡️Laszlo's awesome substack – https://laszlo.substack.com/
- ➡️Laszlo's LinkedIn – https://www.linkedin.com/in/laszlosragner/
- ➡️Laszlo's Twitter – https://twitter.com/xLaszlo
Recommendations:
- 👀Explore/Expand/Extract by Kent Beck: https://www.youtube.com/watch?v=FlJN6_4yI2A
- 👩💻Code Quality – Refactoring by Martin Fowler: https://martinfowler.com/books/refactoring.html
- 📐Geometric Deep Learning by Bronstein/Velickovic: https://www.youtube.com/watch?v=5h6MbQ_65-o
- ✍️Online Writing by Nicolas Cole: https://www.youtube.com/watch?v=Od5J2V-Lmlg
- 📕The Last Shadow by Orson Scott Card: https://www.goodreads.com/en/book/show/7108926-the-last-shadow
- 🎬7 minutes, 26 seconds, and the Fundamental Theorem of Agile Software Development: https://www.youtube.com/watch?v=WSes_PexXcA
🌐Check Out Our Website! https://dagshub.com
Social Links:
- ➡️LinkedIn: https://www.linkedin.com/company/dagshub
- ➡️Twitter: https://twitter.com/TheRealDAGsHub
- ➡️Dean PlbnTwitter: https://twitter.com/DeanPlbn
In this episode, I'm speaking with Lee Harper, Principal Data Scientist at Catapult Systems. Lee holds a Ph.D. in Physical and Theoretical Chemistry. Lee is a teacher-turned-data scientist. We cover the various entry paths into the world of data science, the value of background diversity, security in ML production, and even AI fairness.
Join our Discord community: https://discord.gg/tEYvqxwhah
Timestamps:
- 00:00 Podcast intro
- 01:00 Guest introduction
- 01:39 How did you get into the fields of data science and machine learning?
- 05:04 Coding boot camps vs. academia & diversity of backgrounds in ML
- 09:37 How does the process of bringing your work into production change over the years?
- 13:02 How has the change in the languages used for data science affected production processes?
- 16:01 How do you accelerate the timeframes for getting from POC to production in ML?
- 18:19 Do data scientists reinvent the wheel more often than software developers, and why?
- 22:14 The value of learning how to Google
- 23:00 Recurring themes, challenges, and common issues in data science
- 27:50 Solving for security in ML in production
- 31:57 ML security considerations for startups
- 34:30 Data security considerations in ML
- 35:18 What is the most interesting topic in machine learning right now?
- 38:05 ML fairness, bias, and responsible AI
- 41:44 What does it mean to build a fair or unbiased model?
- 47:15 If you had to choose one challenge in bringing models to production, what would it be?
- 51:00 What are the tools and processes that you use to make the transition to production easier?
- 55:35 About "vendor lock-in"
- 58:00 Your favorite tool recommendations
- 1:03:35 Recommendations for the audience
Relevant Links:
- Linux Command Line and Shell Scripting Bible – https://www.amazon.com/Linux-Command-Shell-Scripting-Bible/dp/1119700914
- Project Hail Mary – https://www.amazon.com/Project-Hail-Mary-Andy-Weir/dp/0593135202
Social Links:
- https://www.linkedin.com/company/dagshub/
- https://www.linkedin.com/company/catapult-systems/
- https://www.linkedin.com/in/leeharper2425/
- https://twitter.com/DeanPlbn
- https://twitter.com/TheRealDAGsHub
In this episode, I'm speaking with Roey Mechrez from BeyondMinds. Roey holds a Ph.D. in Electrical Engineering, with vast experience in computer vision and deep learning research. We discuss the challenges of gluing together infrastructure solutions for an end-to-end ML platform, as well as generating monitoring insights for non-technical stakeholders and combating catastrophic forgetting.
Join our Discord community: https://discord.gg/tEYvqxwhah
Timestamps:
- 00:00 Podcast intro
- 01:00 Guest intro
- 01:49 What does BeyondMinds do?
- 06:24 Audience for an end-to-end ML platform
- 12:14 Communicating with non-technical stakeholders/users
- 15:03 The future of "AI-powered tools", and human-machine collaboration
- 20:04 On complex system orchestration, generating insights from monitoring, and catastrophic forgetting – Biggest challenges in production ML
- 25:23 Why is catastrophic forgetting a hard problem and how do you deal with it?
- 30:02 "Secret" tips on how to get started with automating the retraining process
- 33:30 Generating monitoring insights and observations in a user-friendly format
- 38:12 Making data labeling issues explainable (automatically)
- 45:07 Customizing complex systems per user – Orchestrating an ML platform
- 52:58 API design in ML platform components
- 55:45 Measuring success for researchers, ML engineers, and software developers – can ML work fit into the Agile workflow.
- 1:02:22 Is "time to production" a good metric? Gains in time to production in the real world
- 1:06:02 How do you divide the work between ML researchers and engineers?
- 1:08:39 Recommendations for the audience
Relevant Links:
- A16z blog about AI
- Data Science work in an agile environment – A talk by Dima Goldenberg
- Hayot Kis (Hebrew Podcast) חיות כיס
- Data Engineering Podcast
- ACX Podcast
Social Links:
- https://www.linkedin.com/company/beyondminds/
- https://www.linkedin.com/company/dagshub/
- https://twitter.com/roeyme
- https://twitter.com/DeanPlbn
- https://twitter.com/TheRealDAGsHub
In this episode, I'm speaking with Ran Romano from Qwak.ai. Ran built the ML platform at Wix, and we discuss the various data roles, when organizations should focus on ML infrastructure, solving the hard problems of features stores, and one approach to building an end-to-end ML platform.
Join our Discord community: https://discord.gg/tEYvqxwhah
Timestamps:
00:00 Podcast intro
01:00 Guest intro
01:30 Getting into the world of ML and ML Engineering
02:25 The line between Data Engineer, ML Engineer, and Data Scientist
03:50 The future of data roles – what are the trends?
07:21 The most exciting part about taking ML models into production
09:45 Jupyter notebooks in production (again??)
10:41 Signs that notebook productionization might not work
11:42 Building ML-focused CI/CD systems
15:32 Early days of building out the Wix ML platform
16:22 Signs that you might need to focus on ML infrastructure in your organization, and how to convince other stakeholders.
19:21 What part of the platform that you built are you most proud of?
23:51 Defining a feature store and the training/serving skew
27:24 Onboarding data scientists to using a feature store
33:49 When is it too early to build an ML platform?
35:33 Open source components – What parts of your platform did you choose not to build yourself?
40:16 Qwak.ai – What are you working on currently?
41:07 How do you define an "end-to-end" platform in the case of Qwak
44:25 End-to-end vs. Integrated – Advantages and disadvantages
Relevant Links:
-
Qwak.ai: https://www.qwak.ai
-
Wix ML Platform presentation by Ran: https://www.youtube.com/watch?v=E8839ENL-WY
-
https://www.linkedin.com/company/dagshub
-
https://www.linkedin.com/company/qwak-ai/
-
https://twitter.com/TheRealDAGsHub
-
https://twitter.com/DeanPlbn
-
https://twitter.com/ranvromano
In this episode, I'm speaking with Julien Chaumond from 🤗 HuggingFace, about how they got started, getting large language models to production in millisecond inference times, and the CERN for machine learning.
Join our Discord community: https://discord.gg/tEYvqxwhah
Timestamps:
- 01:00 - Guest intro
- 02:14 - Origin of HuggingFace
- 05:37 - Why the focus on NLP?
- 07:45 - The success of the HuggingFace community
- 13:14 - Reproducing models and scaling for the community
- 18:14 - Enabling large models in production
- 23:14 - How HuggingFace scales so many models
- 27:34 - The biggest challenge HuggingFace solved in MLOps
- 32:02 - How HuggingFace transitions from research to production
- 34:44 - Using notebooks vs python modules
- 38:27 - The most interesting topic in ML production
- 40:10 - Fascinating ML research
- 45:24 - Learning new things
- 51:14 - Something that is true but most people disagree with
- 56:54 - Tips to organize research teams
- 1:00:05 - New features for accelerated inference
- 1:01:35 - Most common use case of HuggingFace
- 1:04:17 - Integrating search algorithms into transformer library
- 1:05:09 - Integrating vision models
- 1:06:06 - Long term business model
- 1:10:55 - Automation and simplification of the process of building models
- 1:13:02 - Support for real-time inference
- 1:14:40 - Recommendations for the audience
Relevant Links:
- FastDS: https://github.com/DAGsHub/fds
- BigScience: https://bigscience.huggingface.co
- https://www.linkedin.com/company/dagshub/
- https://www.linkedin.com/company/huggingface/
- https://twitter.com/TheRealDAGsHub
- https://twitter.com/huggingface
In this episode, I'm speaking with Urszula Czerwinska about her path as a data scientist, the projects she worked on, experiences gained as a data scientist, as well as the challenges she's overcome in bringing her machine learning (ML) into production.
Join our Discord community: https://discord.gg/tEYvqxwhah
Timestamps:
0:00 - Podcast intro
1:15 - Guest intro and how you got into data science
3:48 - Finding your fit – research or industry and when to transition
7:23 - What types of ML projects do you specialize in
10:41 - ML explainability and interpretability
15:26 - ML explainability with non-technical stakeholders
17:13 - What problems does your team solve within the organization
20:56 - ML in production – how to bring your ML projects from research to production
25:17 - The tools you can't live without
28:11 - Do you have a set process for productizing ML projects
30:08 - Team structures and communication for data science teams
33:42 - Who's in charge of setting up infrastructure for a project and job title discussion
36:29 - Interesting tools and repositories you work with
39:30 - How do you stay up to date
42:00 - Biggest challenges for you in ML
45:12 - Favorite and least favorite thing about being a data scientist
49:52 - Handling a workplace that doesn't understand what a data scientist is
53:07 - Data scientists are 🦄 53:30 Good papers you read recently
58:12 - Tips to improve the data science workflow
Relevant Links:
-
flair: https://github.com/flairNLP/flair
-
AllenNLP: https://github.com/allenai/allennlp
-
Papers with Code: https://paperswithcode.com/
-
Dair.ai newsletter: https://dair.ai/newsletter/
-
HuggingFace: https://huggingface.co/blog