← Selected Work

Zamp AI Dashboard

Cutting load time, shipping an architecture for AI agents to drive the UI, and writing the meta-framework that turned 10-day features into 2-day ones.

Role
Lead Engineer · Team of 5
Year
2025 — Present
Stack
TypeScript · SSR · Service Workers · Monorepo
Glowing ring of fine geometric lines

Context

Zamp's dashboard was the canvas a team of AI agents needed to operate on — but it was loading in twelve seconds. The work was equal parts performance, architecture, and developer experience.

What I built

  • 01Reduced load time from 12s to 2s — an 83% improvement — via SSR, pre-caching, Service Workers, and aggressive chunk optimisation.
  • 02Designed a global event-based architecture so AI agents can call into the frontend as function calls or full workflows.
  • 03Wrote a meta-framework with optimistic updates, live sync, prefetching, caching, and declarative data dependencies. A feature that took 10 days now takes 2.
  • 04Migrated to a monorepo, wired up Web Vitals observability, and set the standards the rest of the frontend team works against.

Outcomes

Load gain
83%
From / To
12s → 2s
Feature velocity
10d → 2d
Team
5 engineers

Reflections

An AI-first dashboard reframes the frontend: it isn't only consumed by users, it's an API for agents. That changed almost every architectural decision.