filmov
tv
Craft, Scale, Win: Your ML Platform Journey! // Andrei Vishniakov // Meetup IRL #50 Berlin
Показать описание
MLOps Community IRL Meetup #50! The Berlin MLOps Community talked to Andrei Vishniakov, Senior ML Engineer @ZenML GmbH.
//Abstract
Many companies at some point face a need to have some ML Platform to accelerate ML adoption. How can you get one and make sure your team can maintain it in the long run? This talk will give you a practical guide on how to approach ML Platform building: What options do we have, and what are the drawbacks? How do you ensure the stability of your ML Platform in a world where new MLOps tools pop up every week and make your ML Platform usable by Data Scientists?
// Bio
Andrei is a Machine Learning Engineer based in Berlin with a strong passion for building ML Platforms at scale. Having 10+ years of industry experience in total and 5+ years solely building ML Platform Solutions, Andrei's experience as a "Full Stack Data Scientist" helps a lot to feel platform users' pain points and help solve them. Currently, Andrei is working on making MLOps simple for all at ZenML (a startup from Munich). Previously Andrei worked at HelloFresh, one of the biggest Meal Kit delivery companies in the world, where he was building an ML Platform for Data Science, Data Engineering, and Data Analytics teams from scratch.
// Jobs board
// Related links
Telegram: @avishniakov
----------- ✌️Connect With Us ✌️-------------
Follow us on Twitter: @mlopscommunity
Timestamps
[00:00] Building a platform for users
[00:40] Andrei's background
[02:00] Needing MLOps
[06:33] Options
[08:57] Is best-in-breed a silver bullet
[10:12] Build your platform for users
[12:24] Tip #1: Avoid technical bias to the max
[13:43] Tip #2: Evaluate your ideas
[15:23] Tip #3: Total Cost of Ownership (TCO)
[16:34] All-in-one TCO vs Best-in-Breed TCO
[17:02] Tip #4: Hide complexity
[20:53] Tip #5: Look for support
[23:02] Wrap up
//Abstract
Many companies at some point face a need to have some ML Platform to accelerate ML adoption. How can you get one and make sure your team can maintain it in the long run? This talk will give you a practical guide on how to approach ML Platform building: What options do we have, and what are the drawbacks? How do you ensure the stability of your ML Platform in a world where new MLOps tools pop up every week and make your ML Platform usable by Data Scientists?
// Bio
Andrei is a Machine Learning Engineer based in Berlin with a strong passion for building ML Platforms at scale. Having 10+ years of industry experience in total and 5+ years solely building ML Platform Solutions, Andrei's experience as a "Full Stack Data Scientist" helps a lot to feel platform users' pain points and help solve them. Currently, Andrei is working on making MLOps simple for all at ZenML (a startup from Munich). Previously Andrei worked at HelloFresh, one of the biggest Meal Kit delivery companies in the world, where he was building an ML Platform for Data Science, Data Engineering, and Data Analytics teams from scratch.
// Jobs board
// Related links
Telegram: @avishniakov
----------- ✌️Connect With Us ✌️-------------
Follow us on Twitter: @mlopscommunity
Timestamps
[00:00] Building a platform for users
[00:40] Andrei's background
[02:00] Needing MLOps
[06:33] Options
[08:57] Is best-in-breed a silver bullet
[10:12] Build your platform for users
[12:24] Tip #1: Avoid technical bias to the max
[13:43] Tip #2: Evaluate your ideas
[15:23] Tip #3: Total Cost of Ownership (TCO)
[16:34] All-in-one TCO vs Best-in-Breed TCO
[17:02] Tip #4: Hide complexity
[20:53] Tip #5: Look for support
[23:02] Wrap up