DataFramed #225 The Full Stack Data Scientist with Savin Goyal, Co-Founder & CTO at Outerbounds
8 snips
Jul 11, 2024 
Savin Goyal, Co-Founder and CTO of Outerbounds, dives into the evolving role of the data scientist, blurring lines between data science and software engineering. He discusses the concept of the Full Stack Data Scientist, emphasizing the need to transition models from projects to production. Savin shares insights on the importance of model reliability, the challenges of MLOps, and the power of tools like Metaflow to streamline the deployment process. He also highlights strategies for overcoming hurdles in machine learning implementation and the collaborative future of AI.
AI Snips
Chapters
Transcript
Episode notes
Data Science 'Production' Is Broad
- "Production" in data science is a spectrum, varying by project maturity and organizational needs.
- It can range from deploying models in live A/B tests to informing business strategy with offline analyses.
Production at Netflix
- At Netflix, deploying models to production often involved A/B testing against live traffic.
- Offline model analysis also informed crucial decisions like pricing strategies, similar to Spotify's recent price adjustments.
Not All Models Reach Production
- Many data science models never reach production, which is acceptable due to the experimental nature of the field.
- The key is quickly iterating and improving models.