Nº 001 / ARGUS

ARGUS transforms trading application logs into structured telemetry and actionable alerts. Built for the SMU × UBS FinTech capstone, it uses lightweight agents to monitor distributed trading application instances through a centralized backend.

Diagnosing trading issues required manually inspecting logs across individual hosts. Execution failures, stalled FIX sessions, and performance degradation could go unnoticed until someone investigated.

Telemetry collection also had to operate alongside a latency-sensitive trading application, with minimal resource overhead and without retaining raw logs or full FIX payloads.

Lightweight agents tail FIX 4.4 and application logs, handling rotation and restarts while parsing events through bounded queues and worker pools.

Structured events are aggregated into counters and snapshots. A configurable rule engine evaluates thresholds, while an alert router delivers HMAC-signed callbacks with retry and backoff.

A FastAPI backend centralizes telemetry ingestion, querying, and alert APIs, supported by Redis for shared state. A simulator generates synthetic logs for end-to-end testing.

Led development of the rule engine and alerting pipeline, including configurable thresholds, evaluation loops, FIX session monitoring, and signed alert callbacks.

Built the counter producers, session tracker, and snapshot emitter that supply telemetry to the rule engine.

Integrated most of the team's pull requests into the main branch and groomed product backlog, coordinating delivery across components.

• Python
• asyncio
• FastAPI
• Redis
• FIX 4.4
• uv
• pytest

Validated the complete telemetry pipeline—from log ingestion and parsing to aggregation, rule evaluation, and signed callbacks—using simulated trading activity.

Achieved over 800 automated tests across the workspace.

In progress Natural-language diagnostics through Microsoft Teams, using MCP and Copilot integration.