At Heuronics, I built a browser-based telemetry and control dashboard for remote ROV operation, streaming MAVLink telemetry, sensor data, and mission state over WebSockets. I brought live camera latency below 100 ms by moving the video feed to a WebRTC pipeline.
At House Technologies, I built Python REST APIs and video-processing pipelines integrating ML object tracking. I also designed event capture and alerting for more than 500 telemetry events per minute.
On my Tech-Research Agent project, I built a pipeline that crawls live web sources and returns cited answers, using retrieval and reranking to improve grounding. My other projects include an email outreach platform and an asynchronous vehicle-damage inference pipeline.
I self-host services on Linux and Raspberry Pi, reworked my OpenHD fork to spread its workload across CPU cores, and contributed a pull request to LiveKit Agents. I’m also researching hybrid neural compression for running LLMs efficiently on generic x86 CPUs.
