Skip to main content

Work

Four systems. All shipped. All still running.

No client case studies yet, and I would rather say that than invent them. What I have instead is four systems I designed and built myself. They are all live, so you can open them and judge the work directly.

01 — 04 · all live

Designed, built and deployed solo

01FlagshipLive

Marketing Ops Platform

A multi-tenant platform for running marketing operations on autopilot. Connects the tools an SMB already pays for, then runs scheduled agents on top: morning digest, ad-spend anomalies, review replies, local search reporting. Approval workflows and WhatsApp-first alerts keep a human in the loop.

The hard part

Multi-tenant isolation with per-tenant credentials, and an approval queue that has to stay correct while an agent runs at 3am and nobody is awake to catch it.

  • Next.js
  • Supabase
  • Claude API
  • MCP
  • WhatsApp API

Multi-tenant architecture, agent orchestration across live business systems, and human-in-the-loop approval design.

denver-ai-tech-dashboard.vercel.appMarketing Ops Platform product screenshot
02

Denver Trades

A vertical CRM for commodity exporters built on real customs records rather than estimated intent. Buyers are scored 0 to 100 on commodity match and trade-lane fit, inbound WhatsApp RFQs are parsed into structured demand cards, and deals move through nine trade-specific stages from new lead to shipped.

The hard part

Turning messy customs filings into a buyer score an exporter will actually act on — the model is only useful if they trust the number.

  • Next.js
  • Supabase
  • Vector search
  • Gemini
  • WhatsApp API
03

Ask Data

A production AI BI tool that turns any spreadsheet into a conversational analytics surface. Built in two weeks. React + Vite, Tailwind, Recharts, Claude API via Vercel serverless functions.

The hard part

Keeping the model honest about a schema it has never seen, and refusing the question when the data genuinely cannot answer it.

  • React
  • Vite
  • Claude API
  • Recharts
  • Vercel
04

TradeBook

A production-grade AI app handling real-time data, AI analysis, complex tax logic across two jurisdictions, and an opinionated UI. Built solo in six weeks. Proof I can ship hard things, not demos.

The hard part

Two tax regimes with contradictory rules, computed correctly over the same trade history.

  • Next.js 15
  • Supabase
  • Claude API
  • Edge Functions

One last thing

If you scrolled this far,
you have a question.

Ask it in an email. One paragraph is plenty, there is no form to fill in, and it reaches me directly — there is nobody else here for it to reach.

What happens next

  1. 01You send one paragraphWhat's manual today, roughly how often it runs, and which systems it touches. That's enough.
  2. 02I reply with what I'd actually doA real first take, not a brochure — including when the honest answer is that you don't need AI for this.
  3. 03We talk, if it's worth talking aboutThirty minutes, on your hours. No pre-call form, no discovery deck, no pitch.

Based in India, working across US, UK and UAE hours. Calls run on your clock, not mine.