Microsoft Research CodaLab Case Study | SproutVest
Note on this page. This was a part-time Microsoft role, not a client engagement in the consulting sense. Erick was Principal Program Manager, Artificial Intelligence at Microsoft Research from 2011 to 2015, contracted through SproutVest from 2013 onward. It appears here because the judgment SproutVest sells was built doing this work.
At a glance
Company: Microsoft Research
Role: Principal Program Manager, Artificial Intelligence
Industry: Academic Technology, Open Source Software, Data Science Platforms
Tenure: Part-time, 2011–2015, contracted through SproutVest from 2013
Scope: Product strategy, community building, open-source program management
Headline result: An organically grown research tool turned into a scalable, community-run competition platform hosting 300+ competitions
Context
CodaLab is Microsoft Research’s open-source platform for collaborative data science competitions, used by researchers, academics, and data scientists to host and evaluate machine learning challenges. It had grown the way research software usually grows: organically, feature by feature, without a unifying product strategy. Erick Watson took product direction for it, with a mandate to give the platform a real roadmap and to build the community and sustainability foundations it would need to outlast any single sponsor.
The challenge
Four problems compounded each other. The platform had no unified roadmap, so development priorities shifted with whoever was contributing. There was no community yet: academics, researchers, and open-source contributors used the tool, but nothing existed to onboard them, retain them, or turn them into maintainers. There was no viable model for funding ongoing development within the open-source framework. And growth itself was becoming a liability: more users and more hosted competitions were creating performance and infrastructure bottlenecks.
What Erick did
Erick began with user research across universities, research labs, and open-source communities, and used it to define a product vision and prioritization framework that aligned development with what competition organizers and academic users actually needed. Around that core, he built the community machinery from scratch: contributor onboarding guides, open discussion forums, and recognition programs that turned first-time contributors into invested maintainers.
For sustainability, he developed funding strategies built on academic partnerships, sponsorships, and institutional collaborations, giving the platform a path to longevity beyond Microsoft’s internal support. And with Microsoft engineers, he worked through the scalability constraints, redesigning backend infrastructure to handle substantially more concurrent users and competition data.
Results
Over the tenure CodaLab went from an organically grown research tool to a structured, scalable, community-run platform:
- 300+ competitions hosted on the platform
- A contributor community grown from zero to 3,000+ members
- A unified product vision and roadmap grounded in academic user research
- Infrastructure upgrades that increased competition capacity and stability under high load
- A sustainability roadmap for funding and governance beyond Microsoft’s internal support
The work above starts with an honest read of the leadership behind it
The Trellis Leadership Diagnostic applies that discipline to how you run your company: 24 behavior-anchored items across six dimensions, about 12 minutes, instant results, free to take self-serve. The full version adds Erick's written perspective and a 90-minute debrief; $2,500, credited in full toward the first month of coaching.
Take the Leadership Diagnostic →Exploring a fractional or advisory engagement instead? Book a discovery call →
