Project

George: Your Robot Companion!

Back to Projects

Media

George robot hardware close-up
George in final demo form with integrated expressive hardware.
George robot prototype view
Early integration view while tuning physical personality cues.
Y Combinator office sign during RoboHacks
At the Y Combinator office for RoboHacks weekend.
Spirob research inspiration reference
Spirob-inspired tail concept that informed our motion system.

What is the project

At YC RoboHacks, our team built George, a sensory robot companion that combines Gemini DeepMind VLM reasoning, ElevenLabs voice, and custom expressive hardware to react with human-like personality.

  • I spent the weekend at the Y Combinator office competing in RoboHacks, where our team set out to build a robot that does more than execute commands.
  • We integrated Gemini DeepMind VLM to handle reasoning and contextual decision-making, giving George a flexible AI brain.
  • We used ElevenLabs for lifelike speech output so the robot's responses felt natural and emotionally grounded.
  • On the hardware side, we built a custom physical expression layer on the Innate MARS platform, including expressive eyebrows and wings to communicate internal state.
  • I helped implement a Spirob-inspired tail mechanism with a modified hardware control setup to deliver bio-inspired motion and nuanced behavior.
  • The final system could talk, emote, and physically react in ways that made interactions feel more human than a typical command-and-response robot.

What I learned

  • Power distribution was one of the hardest parts of the build: we had to reliably run a six-actuator expression system while keeping noise and voltage drop under control.
  • I learned how to architect mixed-voltage rails in a robotics prototype, using a DC step-down (buck) converter to regulate a 12V source down to a stable 5V rail for lower-voltage components.
  • We pushed the NVIDIA Jetson Orin GPIO budget hard, so I had to plan pin allocation carefully across actuator control signals and peripheral interfaces.
  • I designed and iterated CAD mounts for the eyebrows and tail mechanism so the expression hardware was rigid, serviceable, and aligned during motion.
  • Implementing the Spirob-inspired tail system taught me how closely mechanical geometry, actuator limits, and control behavior must be co-designed to produce smooth bio-inspired movement.

Skills Learned

  • Human-Robot Interaction Design
  • Embodied AI Prototyping
  • Multimodal Systems Integration
  • Embedded Power Distribution
  • DC-DC Buck Converter Integration
  • GPIO Resource Planning (Jetson Orin)
  • CAD for Robotics Mechanisms
  • Gemini DeepMind VLM Integration
  • ElevenLabs Voice Pipeline Integration
  • Robotics Hardware Control
  • Rapid Prototyping in Hackathons
  • Cross-Functional Team Collaboration