About

The team

Ari (Ariel Anders) — roboticist, MIT PhD in robust manipulation. Previously onboard planning at Waymo and first roboticist at Robust.AI. Currently launching an independent robotics consulting practice.

Sree Dhyuti Nimmagadda — Ari's teammate for the hackathon (dhyutin on GitHub).

Sree and Ari at the Robotics Center of Silicon Valley with the xArm robot

The hackathon

Viam + MongoDB "Physical AI" robot hackathon, October 29–30 2026, at the Robotics Center of Silicon Valley in San Francisco. Two days, one xArm x850, judging on functionality, ambition, creativity, modularity — and 5% Halloween.

xArm x850 holding the lidded coffee cup
Day 1 Pivot to cup idea Swirl macro + recordings Pipeline end-to-end Day 2 Dashboard + docs Simulate backup mode Demo at 3:30 PM

Day one: the team chose the lidded-cup idea, wrote the deterministic swirl macro, and recorded the first experiment matrix — pick, swirl, place, repeated across water, weights, screws, and empty cups. The pipeline ran end to end: torque telemetry → 10-D signature → Atlas vector search → prediction.

Day two: the dashboard, the docs site, the simulated-replay backup mode, and the evaluation bake-offs (hand features vs PCA vs Chronos, kNN vs LDA).

Build On It

This was a fun project tackling an important problem: giving robots a sense of touch from the hardware they already carry. The code is MIT licensed open source. Take it, extend it, and teach your robot how to feel.

Repository: github.com/arii/feels · Dataset: dhyuti-n/feels-cup-dynamics