rumman.ahmed

Sydney, Australia

Rumman Ahmed

Solutions architect at the Powerhouse Museum. I write here about the systems I build and run — event-driven platforms, cloud infrastructure, and AI agents in production.

Concept sketch of the Powerhouse Parramatta EAST and WEST buildings — a diagonal lattice structure raised on angled columns.
Powerhouse Parramatta — EAST & WEST. The estate I help build the digital layer for.

Featured · AI context management

162 tokens, not ~12,891 — 79.6x cheaper to boot an agent.

Booting an agent cold on a new project costs ~12,891 tokens of re-explaining context. With myAI's git-versioned brain, the same boot costs ~162 — measured by scripts/brain_token_eval.py, not claimed. Sessions are commits, wrap-up is a merge, and “brain blame” maps a shipped commit back to the memory behind it. An autonomous runner works a task queue on a free tier, backed by 64 specialist agents, 136 skills, and 134 MCP tools.

  • 162 tokens to boot vs ~12,891 cold — 79.6x cheaper, measured
  • Git-versioned memory you own — sessions are commits, wrap-up is a merge
  • An autonomous runner works a task queue overnight, on a free tier
  • 64 specialist agents, 136 skills, 134 MCP tools

01 / Writing

Recent posts

Working notes on architecture and delivery, written as things ship.

How myAI's brain works · an animated tour

Every AI session normally starts from zero. An animated tour of how myAI gives every model the same memory, tools and to-do list — the 08:00 boot, git-versioned memory, distillation, recall, one gateway, a local runner and a multi-model swarm.

Six Tabs, One Screen · a front-of-house app that owns no data

A front desk answered visitors from six different tools on a shared tablet. Seven weeks later it runs on one screen that stores none of their data: every tile reads its source of record, the network decides who signs in, and the design fails closed. The architecture, the day it broke, and what is not proven yet.

Pitch to Phones · a staff-picks feature from idea to app stores in 27 days

A staff-picks feature for a museum visitor app went from pitch to a production prototype the same afternoon, then from CMS schema to TestFlight and the Play internal track in about thirty hours. Prototype the contract, not the screens: one shape, four teams building in parallel, a source label on every response, and a shared release pipeline with no special path.

02 / Now

What I'm working on

Powerhouse Museum

Head of Solutions Architecture. Content platform, ticketing and identity architecture across the museum's digital estate.

More

myAI

An after-hours experiment in multi-agent engineering — a task queue worked by scheduled agent sessions. This site ships through it.

More

MBA (Technology), UNSW AGSM

In progress.

03 / Work

Selected work

A few of the systems I've worked on, with the architecture written up.

Powerhouse Museum2025 — now
Architecture across the Powerhouse renewal — content platform, ticketing, identity and exhibition technology for Australia's museum of applied arts and sciences.
ArchitectureStrategyAWSHeadless
Powerhouse Museum2025 — now
A headless content platform built around an event backbone — CMS, collection and DAMS changes fan out to search, cache and every consuming channel without any system polling another.
Event-drivenHeadlessAWSNode.js
Australian Broadcasting Corporation2017 — 2025
Platform modernisation across ABC digital — cloud-native broadcast infrastructure, FAST channel automation, enterprise search, and live event architecture for major national broadcasts.
AWSMediaEvent-drivenDevOps

04 / Contact

Let's build something that holds

I'm always up for a conversation about platform architecture, museum and media technology, digital transformation, or where agentic AI is actually useful. Reach me directly: