Vasudev Katariya.

Built an AI symptom analysis engine with weighted scoring & medical ontologies, surfacing accurate diagnoses in under 2 seconds.

Developed an OCR/NLP pipeline (Tesseract + spaCy) to auto-extract structured data from medical PDFs/images, eliminating manual entry.

Integrated Google Maps geolocation to recommend nearest hospitals within a 15 km radius with real-time context.

Secured the platform with JWT auth and optimized MongoDB schemas with indexing, cutting query latency by 40%.

Built MelodyBot, a Gemini 2.5 Flash-powered voice assistant handling natural-language music queries with sub-second response.

Integrated face-api.js for real-time facial emotion detection across 8 moods, auto-curating playlists without user input.

Combined node-cache (server) and a custom LRU cache (client) to slash redundant YouTube API calls and cut music search latency significantly.

Architected a React Portals PiP overlay with inline drag/resize and a persistent streaming context, ensuring zero media disruption across page transitions.

Built an enterprise zero-trust authentication platform preventing account takeovers and insider threats, finishing in the top 10 teams.

Engineered hardware-bound auth using WebCrypto (P-256 ECDSA) and face-api.js facial recognition with a 3-strike fallback cascade (Face ID → PIN → TOTP).

Designed a zero-storage biometric model, keeping facial descriptors and private keys in client-side IndexedDB only, never on servers.

Implemented multi-party authorization via 2-of-3 Shamir Social Recovery and a Duress Canary PIN that silently alerts security under coercion.