Microsoft Research Human-in-the-Loop Data Collection Apps
Problem Space
Build Facebook-integrated apps for crowdsourced ML data labelling (human-in-the-loop). Create dataset visualisation and annotation tools for Kinect, Photosynth, and facial/body tracking research. Deliver MLOps-style pipelines on Azure Cloud Services. Build Silverlight MVVM apps on Windows Phone 7 for sensor-driven input. Work supervised by Pushmeet Kohli (now VP at Google DeepMind).
Architecture & Patterns
- Facebook-integrated apps for crowdsourced ML data labelling (human-in-the-loop)
- MLOps-style pipelines on Azure Cloud Services
- Dataset visualisation and annotation for Kinect, Photosynth, facial/body tracking
- Silverlight MVVM apps on Windows Phone 7 for sensor-driven input
- Supervised by Pushmeet Kohli (now VP at Google DeepMind)
Tools & Stack
Azure Cloud Services, Silverlight, ASP.NET MVC 3, Razor, SQL Azure, Facebook C# SDK
Business Outcomes
- Crowdsourced ML data labelling apps deployed via Facebook integration
- Dataset visualisation and annotation tools for Kinect, Photosynth, and facial/body tracking research
- MLOps-style pipelines on Azure Cloud Services for research data processing
Reusable Narrative Snippets
Built Facebook-integrated apps for crowdsourced ML data labelling (human-in-the-loop) at Microsoft Research, supervised by Pushmeet Kohli (now VP at Google DeepMind), with MLOps-style pipelines on Azure Cloud Services.
Created dataset visualisation and annotation tools for Kinect, Photosynth, and facial/body tracking, alongside Silverlight MVVM apps on Windows Phone 7 for sensor-driven input.
Source Notes
- Derived from role responsibilities and achievements in
config/madu_profile.json; reconciled with JobVia export (madu_alikor_export.json). - Confidence: high