At IntuigenceAI, I built an end-to-end parser that turns engineering diagrams into knowledge graphs, lifting symbol identification from about 30% to 90%. I also engineered an agent harness around a self-hosted open-source 27B VLM that reached 90% on real customer P&ID questions without fine-tuning.
At a stealth startup, I built and own a HIPAA-compliant remote physiotherapy platform, including real-time biomechanical assessment and a multi-agent clinical AI system that accelerated clinician review by 75%. My research has also covered dropout-risk prediction, photovoltaic optimization, and computer vision, including published work on medical imaging, ed-tech, OCR, and object detection.
