Fintech
Signal Backtesting Engine
Wallex — event-driven fintech infrastructure
- Role
- Backend engineer (Wallex)
- Period
- 2022 – 2023
- Scale
- 5M+ events/day
Overview
Signal-backtesting infrastructure at Wallex built on Apache Kafka and the Microsoft Orleans actor model: market data lands on partitioned topics, and each symbol is a virtual actor that fans out to Redis (hot state), SQL Server (positions and history), and dashboards.
This is the same Orleans runtime discipline visible in the Crypto BI codebase (Crypto.Runtime.Grains) applied to signal backtesting: the engine replayed the entire set of trading signals, one pass at a time, checking each in live-time against point-in-time market data — the market exactly as it stood at that moment — to validate the signals rather than execute live trades. Thousands of signals at millisecond latency, 5M+ events processed daily.
- crypto (Runtime.Grains)the authored Orleans actor codebase
- risk-signalsignal/risk service (hamgit)
Tech stack
Core
- C#
- Microsoft Orleans
- Apache Kafka
State & storage
- Redis
- SQL Server
Delivery
- Azure DevOps
- Kubernetes
Design patterns
Event-driven pipeline
Kafka topics (signals.raw → signals.enriched → validation results) decouple producers from the actor cluster.
Actor model
One virtual actor per symbol serializes state transitions without locks — Orleans grains distribute across the cluster.
Hot/cold path split
Redis serves hot state to dashboards; SQL Server keeps the durable position/history record.
System design
