// research
Research
I work on robotics, perception, and how machines read the people in front of them. Right now that means multimodal AI for autonomous characters.
// primary focus
Efficient Multimodal Tokenization for Real-Time Social Perception in Autonomous Robotic Characters
How does a robot read human social cues in real time? The hard part is fusing vision, audio, and balance data into one stream that's light enough to run on edge hardware like the Jetson Orin, while staying fast enough that the interaction still feels natural.
// ballbot build log
Build Progress
Build notes on the mechanical design, control systems, and behavior layer for a self-balancing inverted pendulum robot that interacts with people in character.
// papers & writeups
Multimodal tokenization for real-time social perception
Edge-optimized transformer architectures for robotic perception
Character behavior modeling in autonomous social robots