Experience

Founding AI engineer and sole ML contributor for a medical document intelligence startup.

• Structured unstructured clinical data using OCR, IDP, and LLMs, outperforming Donut Doc VQA and Microsoft LayoutLM on internal benchmarks.
• Cut image-processing latency from 25s to 0.8s via GPU acceleration and async batching.
• Built real-time healthcare chatbots with streaming speech-to-text, and a doctor-patient transcription pipeline producing SOAP notes for EHR integration.

Volunteered on the Wildepod project.

• Fine-tuned computer vision models (Megadetector, ResNet) to automate camera-trap image management and wildlife classification for conservation research.