I build full-stack & AI systems used by 12,000+ people.

// problem

Students had no single platform — everything was fragmented across dozens of WhatsApp groups and manual processes.

// impact

Adopted as the primary campus platform within one semester. Redis caching cut database load by ~70% (measured via pg_stat_statements), keeping responses under 200ms at peak. Ships from a single codebase to web, native Android, and an Electron desktop app, with a GitHub Actions pipeline running Jest unit tests and Playwright E2E tests on every change — run and operated by one person alongside a full course load.

// problem

India's gig workers are excluded from formal credit because delivery/rideshare platforms issue no portable, verifiable income record.

// impact

End-to-end pipeline: upload → OCR (tesseract.js) + per-platform parsers → optional Groq LLM extraction → user review/correction step → aggregated dashboard → SHA-256-hashed shareable PDF with a public verify link that detects edits after issuance. Google OAuth + email OTP auth, Supabase Postgres with row-level security, and a node:test suite that passes on real OCR fixtures. Account Aggregator Phase-2 stubbed.

// problem

Citizens had no easy channel to report civic issues or understand local ward-level governance and representative data.

// impact

Citizens can file an issue against their actual ward and see who represents them — resolution status is public, so follow-up no longer depends on knowing the right office. Admins triage through a dedicated moderation workflow. A Vite + React SPA backed by Supabase (Postgres with row-level-security policies and Edge Functions), deployed on Vercel; ward-boundary data loaded from a GeoJSON dataset and rendered with MapLibre GL.

// problem

Land-record errors block loans, sales, and inheritance for millions of Indians — but discovering what is wrong requires navigating opaque government portals.

// impact

Live automation of the real Karnataka RTC portal (District → Taluk → Hobli → Village → Survey cascade, verified against the production site), privacy-safe AnyROR PDF parsing (uploaded files deleted after parse), a manual-entry fallback, and printable application packets. Evolved from an earlier blockchain concept after field research showed workflow — not data integrity — is the real gap.