The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) Agile Applied AI Research with Parvez Ahammad - #492
9 snips
Jun 14, 2021 
Parvez Ahammad, Head of Data Science Applied Research at LinkedIn, shares his insights on organizing data science teams for success. He discusses the balance of long-term project investments while navigating the challenges of experimentation. Parvez also delves into the impact of differential privacy on member data and the ambitious launch of the GreyKite forecasting library. The conversation highlights the dynamic relationship between applied research and engineering in AI, emphasizing the need for strategic alignment and effective team dynamics in driving innovation.
AI Snips
Chapters
Books
Transcript
Episode notes
Team Positioning and Prioritization
- Parvez Ahammad's applied research team at LinkedIn sits between research and product, focusing on impactful, cross-business problems.
- They prioritize projects with broad utility or strategic importance, like experimentation and privacy-preserving products.
Portfolio Management
- Treat your team's projects like a portfolio of investment theses, similar to Y Combinator.
- Evaluate projects in stages, allowing for adjustments and pivots based on learnings.
Explainability Project Evolution
- Parvez Ahammad's team started an explainability project focused on internal "go-to" market use cases.
- This project generalized well, leading to its expansion and integration into LinkedIn's AutoML framework, ProML.
