Lav Kush

Turns revenue events into recovery cases, scores context, generates LLM recommendations, and enforces deterministic policy & stopping rules before executing actions. Uses mock Razorpay provider for demo repeatability, persists outcomes with Prisma/PostgreSQL, and answers: how much revenue did the workflow actually recover. Demo mode runs 4 deterministic scenarios with metrics, funnel, and attribution dashboard.

Full-stack route computation engine with Express backend and frontend. Computes viable train connections from local timetable/database to minimize transfers and time, using RailRadar API sparingly for validation/enrichment due to limited allowance — not per-train API calls. Focuses on data processing, optimization, and practical UX for real railway data.

Customers discover nearby stores, browse products, reserve items, and request unavailable goods; shopkeepers onboard stores, manage inventory, and receive requests. Built with HTML/CSS/JS frontend and Python + MySQL backend. Includes store registration, product management, booking, and direct contact. Responsive, database-backed, production-minded.

Parses VCF v4.2 with genotype-aware logic (FORMAT/GT), infers star-allele diplotypes, maps to phenotypes per CPIC, assesses drug risk, and generates clinical narratives via Gemini 2.5 Flash with rule-based fallback. Shared engine used by both Express server and Netlify Functions. Honest Unknown states, no fabricated data, XSS-safe rendering, export, dark mode. Built for RIFT 2026 hackathon.