Projects

• 2023 — now

  Robot Teleoperation Interfaces

  Google DeepMind

  Interfaces and infrastructure for operating robots directly — the control surface a person uses to drive real hardware, plus the systems that record those sessions as training data.

  The design problem is that a teleoperation session serves two audiences at once. The operator needs low latency, legible state, and controls that map cleanly onto what the robot can actually do. The dataset needs consistency, clean episode boundaries, and demonstrations that represent the task rather than the operator's habits. Optimizing hard for either one degrades the other.

  • Python
  • C#
  • Unity
  • Teleoperation
  • Data collection
• 2016 — 2019

  UX Research Simulation

  Research institute of an automotive manufacturer · C#, Unity, Processing

  Experimental UX functionality for VR applications used in vehicle design, integrated into robotics work. Researchers could put someone inside a design before it existed in physical form and watch how they behaved in it.

  Alongside the simulation itself, I built tooling to capture research metrics during runtime — the part that turns a compelling demo into something a study can actually draw conclusions from. I also ran demos for both public and internal audiences.

  • C#
  • Unity
  • Processing
  • VR
• 2016 — 2019

  Tango — Augmented Reality Platform

  Google · C#, Unity, Tango

  Contributed code samples and usage demos to the Tango Unity SDK releases — publicly distributed code showing developers what the platform could do, including mesh occlusion and ground detection.

  Writing sample code for other engineers is its own discipline. It has to be simple enough to read in one sitting and honest enough that it does not fall apart the moment someone adapts it. Most of the work was deciding what to leave out.

  • C#
  • Unity
  • Tango
  • AR

Some work is not listed here because it is not public. Happy to talk about it in general terms — get in touch.