At MACHANI ROBOTICS, I architected a production multimodal AI agent using Google's Gemini Live API, with bidirectional audio/video streaming, persistent sessions, and MCP tool orchestration. I also led investor- and client-facing robot demonstrations that contributed to the company's first seed investment.
I owned Kogan, the robot's real-time perception and identity platform, and led its GPU-to-OpenVINO migration for Intel edge inference. I also productized a contactless vitals pipeline and integrated it with the robot's agent ecosystem.
Previously, at LIVESITTER, I built computer-vision systems for retail fraud detection and production inference across 350+ RTSP streams. At ABLELYF, I developed computer-vision and voice-interactive AI systems for autism therapy, and I continue exploring computer vision for edge and robotics through projects such as a GGML YuNet loader.
