Full-stack systems you can trust.
Full-stack engineer in React, Next.js, TypeScript, Go and Python. Four flagship systems with live, interactive architectures: JñānaSetu, ViduraNīti, Flasio and VarunaSetu.
Flagship systems
- JñānaSetu: A multi-tenant RAG platform: self-hosted embeddings and Qwen3-4B answer from a client’s own documents and show where every answer came from.
- ViduraNīti: A spending governance platform for autonomous AI agents: budgets, ML risk checks, human approval and verified x402 + Algorand settlement.
- Flasio: Flask simplicity. Async power. An async-first ASGI web framework I wrote from scratch and published on PyPI.
- VarunaSetu: A physics-informed, spatially aware research prototype that verifies rooftop rainwater-harvesting telemetry before it is stored or credited toward water allocation.
Skills, with receipts
- Go: A modular API where identity, ingestion, retrieval, inference and chat are separate packages in one container, with an inference adapter that enforces limits, timeouts and cancellation.
- React · Vite · TypeScript: An admin dashboard and an iframe-isolated chat widget built as separate Vite apps, plus a ViduraNīti front end designed around TanStack Query and an IndexedDB cache that is never the authority for payments.
- Python backends: A modular FastAPI backend with ten business domains, and Flasio, an ASGI framework I wrote myself and published on PyPI.
- Data and search: Postgres with row-level security and pgvector HNSW search at 384 dimensions, and PostgreSQL as the financial source of truth with Redis kept strictly as cache and queue.
- Spatial data and verification: A stateless Go verifier that checks telemetry against physical bounds and H3 neighbours, an escrow state machine for disputes, and a SHA-256 hash-chained ledger with Ed25519-signed checkpoints.
- Security by design: Server-resolved tenants, private storage, idempotency keys, immutable decision records, and payments marked settled only after on-chain confirmation.
- AI in production: Self-hosted embeddings and Qwen3-4B on llama.cpp for grounded answers, and an ML anomaly score that escalates risky payments to a human instead of deciding alone.
Stack
- Languages: Python, JavaScript, Java, Go, C, C++, Bash
- Frontend: React, Next.js, HTML5, CSS3, Tailwind CSS, Bootstrap
- Backend: FastAPI, Flask, Django, REST APIs, WebSockets, gRPC
- Databases: PostgreSQL, MongoDB, Supabase, Neo4j Aura, Redis
- Cloud & DevOps: Docker, Git, GitHub, Vercel, Render, Kafka, DigitalOcean
- Auth & Security: JWT, Google OAuth, OAuth 2.0, HTTPS / TLS / SSL
- Tools: Linux, Postman, cURL, Pandas, OpenPyXL