Bright.AI
AI Intern Summer 2026 · contract extended through SeptemberI built software that brings machine intelligence into the real world, from overhead cameras on an industrial floor to a wearable for field technicians. After the summer, my contract was extended through September so I could keep developing the computer-vision model I had built from scratch.
- Built an end-to-end computer-vision system for a waste-to-energy facility from raw, unlabeled footage: frame ingestion, SAM3 auto-labeling, and YOLO11 fine-tuning.
- Deployed the detector with TensorRT to a Jetson Orin edge hub (155 ms/frame, zero dropped frames), ran it live in shadow mode, and demoed it to the client and our CTO.
- Benchmarked on-device LLMs for an industrial wearable's knowledge-graph RAG assistant, using a 377-question golden set and a latency model that drove the model-size decision.
- Indexed a 67-repo IoT platform into a queryable code knowledge graph over MCP, then ran an AI-assisted security audit that surfaced critical findings.