Builds

Built, not described.

Each one started with a real question. Here's what I built to answer it.

Dashboard · built with Claude

A live election dashboard, built while the votes were being counted

Results day for the West Bengal election. Counting was live, and the numbers kept moving.

I asked Claude to turn the counting data into one view. Five minutes later I had it: who leads where, by how much, and which seats could still flip.

  • Lead and margin for every reporting constituency
  • Seats at risk: leads under 2,000 votes that could still flip
  • How settled each lead is, based on how much of the count is done
  • Overall counting progress across the state
AI assistant

A personal AI chief of staff, built by a non-coder

My AI assistant knows everything about me. Emails, health data, calendar, notes.

Every morning, 30 minutes after I wake up, it sends me a brief. It nudges me based on my to-do list and tracks my habits. I talk to it the way you talk to Siri or Alexa.

I vibe coded it with AI, so I can change anything about it. No work data ever touches it.

  • Emails I need to reply to
  • How well I slept, plus my readiness score
  • Today's to-do list and meetings
  • Finance and AI news that matters to me
Python · ValuationWalkthrough coming soon

Earnout valuation with Monte Carlo, explained

An earnout is a promise: pay more later if the business hits a target. The question is what that promise is worth today.

This dashboard answers it three ways, side by side: an option pricing model, a scenario-based model, and a Monte Carlo simulation running 10,000 paths by default.

It comes with a written guide, so you can follow the logic, not just the output.

  • Option pricing model (Black-Scholes)
  • Scenario-based model
  • Monte Carlo with a fixed seed, so results are reproducible

Want help applying ideas like these to your own work or career?

Book a 1:1 session →