Founder/CEO at SoftSensor AI and LMDMax. I build practical AI systems—combining vision + LLMs—for:

• Document Intelligence (OCR/HITL, GxP/regulatory, Equity docs, Pharma docs, and multiple flexible document inputs/validation)
• Medical CV (WSI pipelines, grading, decision support)
• Fleet damage detection (high-throughput, structured outputs)

Bias to action. Ship value, measure impact, iterate.

• Document AI platform: OCR + layout + HITL for pharma & insurance master-template checks, provenance).
• Pathology tooling (PRR.ai): WSI pipelines (grading/triage), vision-LLM workflows, reviewer-first UX.
• LMDMax: scalable CV pipeline processing tens of thousands of images/day with human-in-the-loop QA.

• Robust HITL UX: side-by-side PDF/image, extracted fields, evidence trails, reviewer shortcuts.
• Low-VRAM inference without precision loss; boundary-aware metrics (Dice/Jaccard) and A/B configs.
• Multi-agent search & summarization for regulated knowledge bases (auditability, redaction, chain-of-custody).
• Clean MLOps & data contracts: schema/versioning, eval dashboards, reproducible experiments.

• I mentor FIRST Tech Challenge (FTC) teams.
• I also mentor TiE entrepreneurs and young teams on product definition, GTM, and AI MVPs.

• DataIQ 100 (2024) recognition for influence in Data & AI.
• End-to-end OCR/HITL for long-form reports (fraud checks, range/legibility, sequencing).
• High-volume vehicle damage pipeline with structured outputs (bbox, severity, maintenance).

Python, TypeScript, FastAPI/Next.js · PyTorch, ONNX, OpenCV ·
Weaviate/Qdrant, BM25 + BGE rerank · Azure-first (Functions, Storage, App Service, AKS) · Airbyte/DBT/Postgres.

• Practical HITL for regulated OCR
• Boundary-aware metrics for medical CV
• Agentic pipelines that are actually debuggable

• Need a production-minded AI POC or MVP in documents or medical CV?
• Entrepreneurs (incl. TiE): happy to advise on scoping, tech choices, and MVP acceleration—reach out.

📫 Contact via GitHub issues or the contact link on my org sites.