Fintech product, built in code — not just in slide decks.
I'm a product person at FinBox who ships. Most PMs write the spec; I write the spec, then the engine underneath it. My work sits where lending infrastructure, risk decisioning, and AI systems meet — and I judge every idea by the same standard: does it survive contact with production?
🧠 RuleMind AI — an open-source, source-agnostic decisioning engine. Plug in any JSON source (bureau, bank, GST, device, KYC), then author variables, visual rules, scorecards, WoE models and multi-step policies from a dashboard — and promote them dev → uat → prod with full audit trails and rule-level explainability. Ships with 230+ Excel-compatible functions, ML model hosting, multi-tenant isolation, and SDKs for Android, Flutter, JavaScript and Python. Apache 2.0 — fork it, self-host it, extend it.
💛 Couple OS — an AI relationship co-pilot built on Claude Code. 16 slash commands for date ideas, trip planning, surprise planning, milestone letters and streaks, with shared memory via Notion. Runs entirely locally, and gets sharper with every piece of feedback.
⚡ In the workshop — PulseTrade (Kotlin app + Python worker on Fly.io), Trinetra (TypeScript backend), and a fintech PM portfolio at tanmay-fintech-pm.lovable.app.
- Risk is a product surface. Scorecards, policies and decision traces deserve the same UX rigour as a checkout flow.
- Explainability isn't optional. If a decision can't be traced back to a rule, it shouldn't be in production.
- Prototype > proposal. The fastest way to align a room is a working demo.
- Open by default. Good infrastructure should be forkable.
Python · TypeScript · Next.js · FastAPI · Kotlin · Postgres · Redis
Docker · Fly.io · Playwright · Claude Code · Notion API
LinkedIn · Portfolio · 🌏 India (UTC +5:30)
Always up for a conversation about credit decisioning, AI agents that actually do work, or why your rules engine needs an audit log.