Skip to content
Aarav Jain
All work

Hack-Nation 7th Global AI Hackathon, Oct 2026

ExpertAI

An AI apprentice for any desk job.

1 min read

My role
Led the build and wrote most of the code: the Next.js app, apprentice panel, Work Map, jobs dashboard and API routes. Made the launch film and played the expert in the live demo
Where
Hack-Nation 7th Global AI Hackathon
When
Oct 2026
Stack
  • Next.js
  • TypeScript
  • Claude (vision + LLM)
  • ElevenLabs Agents
  • Supabase
  • Chrome extension
  1. The problem

    Preflop

    Every desk job runs on rules nobody wrote down. The goal: your best people, on every shift.

  2. The approach

    Flop

    ExpertAI watches an expert work on screen and stays quiet while they’re busy. When they pause, it asks by voice why they did that. Every answer becomes a rule on a Work Map, traced back to a quote.

  3. The hard part

    Turn

    Knowing when to ask. It waits for 1.5 seconds of quiet across screen, voice and input, and only asks about what the screen can’t explain. About 24 hours of build time, with every package written against shared contracts.

  4. What shipped

    River

    A platform, not a demo: each job is a JSON profile, so the same engine runs a returns desk or invoice approval. A Chrome extension puts it on any web app and pauses a risky save with the expert’s rule and quote.

The result, or as poker players say, the showdown

A new hire gets stopped before the wrong refund, in the expert’s own words: “That one’s over $100. Let’s get the shift manager.”