
Alpine Valley: Berry-Harvesting VLA
Founder in Residence / Latent Robotics
Building an agri-robotics company led me to the deeper problem: controllable embodied AI.
I build robots and AI systems that survive contact with the physical world.
Zero-to-one bets: a berry-harvesting company, a voice app I shipped to both app stores, and a string of hackathon prototypes. Proof that I actually finish and ship things.
Grasping and contact-rich control that still works when objects shift, slip, and refuse to cooperate. Learned policies backed by classical control so they stay reliable.
Vision and state estimation that hold up in bad light, occlusion, and clutter, from stereo depth to figuring out what a policy is actually seeing.
Robots that have to work outside the lab, on strawberry rows, uneven ground, and port terminals, where things break in ways you never see indoors.
Keep the reliability and fine control of classical robotics while gaining the generality of foundation models.
Knowing what a policy does is not enough. Expose handles inside the task and use them to steer it.
Robots earn trust when people can inspect, correct, and step in. That loop is where reliability lives.

Founder in Residence / Latent Robotics
Building an agri-robotics company led me to the deeper problem: controllable embodied AI.

Hugging Face technical blog
A supervision layer around a SmolVLA policy, built to make the rollout easier to inspect and steer.

Open-source robotics tooling
My first serious open-source robotics contributions.
Qafka Robotics
Lely
TU Delft
NITK Surathkal
Hugging Face
IEEE
Frontiers
Hackster
Snap
Junction
Hugging Face