Engineering Manager | AI-Native Healthcare Architect | Oncology EMR | Fitness AI

github

Tushar Sarang  • Peerlist

<a target="_blank" rel="noopener noreferrer" class="break-all link-break text-gray-500 hover:underline" href="http://De-ID.org">De-ID.org</a> is a browser-based HIPAA PHI de-identification engine built for healthcare AI teams, researchers, and privacy-first developers.Most AI tools require you to upload sensitive medical documents to their servers. That creates compliance risk, audit complexity, and trust issues.De-ID solves this differently.It runs ONNX-based NER models directly inside the browser using ONNX Runtime Web. No PHI leaves the device. No server storage. No retention.Upload a clinical PDF → extract structured text → detect all 18 HIPAA identifiers → download anonymized output as Markdown, JSON, or TXT.Key capabilities:• Client-side ML inference (true browser execution) • PDF structure preservation (headers, tables, formatting) • Entity confidence scores • Rule-based cascade for SSN, MRN, regex patterns • Progressive Web App deployment • Zero infrastructure cost for privacy-first modeFor enterprise use cases, a cloud version powered by FastAPI + ONNX Runtime enables batch processing, audit logs, and API integrations while enforcing strict no-retention policies.De-ID is built for:• HealthTech startups building LLM features • FHIR and EMR developers • Clinical NLP researchers • Compliance-conscious AI buildersIf you're building healthcare AI and need PHI stripped before sending data to LLMs, this is the missing layer in your stack.