Global bank
Risk engine rebuilt. 40× faster close.
A monolith handling $2T in positions re-architected into a streaming risk platform — overnight batch became a four-minute continuous close.
4 min
continuous close (was 9 hours)
0
missed opens since launch
$2T
positions under management
Timeline
11 months, discovery to full cutover
Built with
- Algoryq Grid
- Kafka
- Rust
- PostgreSQL
- Kubernetes
The problem
The bank's overnight risk batch had grown to nine hours. A failed run meant trading opened blind — and runs failed monthly. Regulators had noticed.
What we built
We re-architected the monolith into a streaming platform on Algoryq Grid: positions revalued continuously as market data arrives, with the batch system kept live in shadow until parity was proven over a full quarter.
