Hit the (ML) Road(map) Jack
summary
A concrete framework emerges: generate a large list of one model task ideas, score each by five concrete metrics (model development, performance measurement, customer value, time to market, engineering), and schedule work in a six‑step research cycle (literature review → dev → supervised experiment → real‑world experiment → push to production → monitor). Rituals include research review meetings, tech one‑on‑ones, weekly spotlights and project scrums. A single cycle typically takes seven weeks, and the emphasis is “done better than perfect.”
AI-written summary · about
chapters (8)
- 0:00Introduction
- 3:59Prioritized ML tasks
- 17:53Research cycle framework
- 25:39Team rituals for research
- 36:31ML vs software lifecycle
- 38:57Speed versus research quality
- 41:06Time frames for cycles
- 43:14Final ML roadmap
related talks
29mDriving clarity in ambiguous projects
15mResearch to Reality: Bringing Frontier ML Research to Production
27mAn Engineering Leadership Roadmap
29mLeveling Up from Planning to Production
58mProject Management for Engineers
28mJourney from Design Afterthought to Design First
48mBuilding a Product and Engineering Relationship that Works
30mMy first 90 days, kick starting a new team for long term success