This week goes out the first edition of the slack weekly newsletter! Subscribe here and you will never miss abeat in the community!
Upcoming Meetup
Building a Modern Data Analytics Stack
Jeff Katz has taught and developed curricula in machine learning, web development, and data engineering since 2015.
In 2019 he co-founded Jigsaw Labs, a part-time boot camp that trains students to become data engineers in six months.
In this Meetup, Jeff will set up a modern data analytics stack with Fivetran, DBT, and Snowflake among other tools. Jeff will also show how to quickly set up a modern data analytics stack so that companies can collect data from their frontend and backend, move to an analytics database, and then perform queries to ultimately present in a data dashboard.
Don't listen
unless you are doing ML in production
For this coffee session, we talked to Kyle Morris, Co-founder of banana.dev and former software engineer at Cruise AI.
Key takeaways: - Adaptability
is more important than skills, skills can be acquired and learned. - Everything can be reduced to 3 axes when serving things: latency, cost and scalability/reliability. Reliability is expected. - Focus on ONE thing and be 10X better at that. - If you can't find someone willing to pay top dollar to what you are doing, you are doing something wrong. - Paying for idle GPUs is really dumb. - Hurdles in productionalizing ML can be limited by learning 3 things initially: Docker, Kubernetes and Infrastructure as Code (Pulumi).
"I want to make sure I don't regret not pushing hard
enough"
Todd is a Director at Google and Machine Learning for Site Reliability Engineering Lead. He is also a Site Lead for Google’s Pittsburgh office. Being a googler for over 13 years, he's now actively sharing his wisdom with the Community.
Why Companies Are Switching to Flyte for Their ML & Data Orchestration Needs
Flyte is a workflow automation platform, enabling highly
concurrent, scalable and maintainable workflows for complex, mission-critical machine learning and data processing at scale. Conceived at Lyft, Flyte was open-sourced in early 2021 under LF AI & Data, and was recently promoted to a Graduate Project in Jan 2022.
Flyte’s most notable features include it being:
Open-sourced and Kubernetes-native
Built around a core unit of execution (task), which is easily shared and reused
Language-independent, yet type safe
Extensible (backend & SDK’s), scalable, and battle-tested
Early users and contributors to Flyte included Spotify and Freenome, which had rapidly expanded to span across various software development industries, namely 3D mapping, service and delivery platforms, autonomous vehicles, in addition to Biotech and AI. Each company was able to find that sweet spot where Flyte would seamlessly fit into their ML
infrastructure, drastically reducing processing time and enabling scalability while requiring minimal integration efforts from the data science teams. Flyte has proudly been named one of the Top 10 Python Packages for MLOps this year.
