We sat with their team, walked every existing flow, and wrote the brief on day three.
The case study
From a 90-day windowto a body of work.
A re-architecture from monolith to a typed event-sourced core, with a trading console the energy team actually uses at 3am. Cut median pricing latency from 1.2s to 47ms. Below is the long version — the brief, the approach, and the numbers we put on the wall after launch.
01 · The challenge
Helio's pricing was running on a Rails monolith that tickled a single Postgres primary. The 3am trading desk was a Discord channel and a prayer.
02 · The approach
We carved the hot path into a Go + Kafka event-sourced core, added a typed Rust optimiser, and rebuilt the trader console as a single screen with a typed data layer. Every screen has an undo, every action is auditable.
03 · The outcome
47ms median pricing latency, 9 GW of renewable capacity traded live, and a trading desk that finally sleeps. Pager volume fell 71% in the first month.