Trading & Risk Systems

Problem

Commodities and futures trading software has almost no tolerance for lag or ambiguous state — a hedging dashboard that's slow or a notification that arrives late is a dashboard that's wrong.

Engineering challenge

At Quantorix, the API layer was too slow under real load and a trading service needed to scale from one region to global usage without downtime. At Phlo Systems, the notification layer for a risk-management platform needed to be decoupled from the main API so market-data bursts wouldn't degrade the rest of the system.

What I built

• Quantorix: replaced a repository-pattern data layer with optimised controllers, cutting P90 API latency from ~500ms to ~150ms at 60 requests/second
• Quantorix: scaled a real-time trading service from a single region to 114 countries using SignalR with backpressure handling, with zero downtime during rollout
• Quantorix: integrated NLP-powered search over the ELK stack to improve intent recognition on trading queries
• Phlo Systems: architected a real-time notification engine on SignalR and Azure Communication Services as a standalone service, off the critical path of the main API
• Phlo Systems: built fault-tolerant reporting pipelines on Azure Functions and led the React 16→18 and .NET Core 6→8 migrations

How it works

Outcome

A 70% cut in P90 latency and a global rollout with no downtime at Quantorix; a decoupled, fault-tolerant notification and reporting layer at Phlo Systems.

Stack