Finance Analytics Dashboard

The problem

Finance teams were processing corporate card statements by hand — slow, error-prone work that produced little insight. The data existed, but turning it into answers took days of manual effort each cycle.

The approach

• Statement ingestion pipeline — uploads are parsed and normalised asynchronously with ARQ workers behind Redis.
• AI classification — transactions are categorised by LLMs, with human-in-the-loop review for low-confidence cases.
• RAG-powered agents — analysts ask questions in natural language and get grounded answers over their own tenant’s data.
• Multi-tenant architecture — strict isolation between client organisations across data, jobs and dashboards.
• Containerised deployments — reproducible Docker-based environments from development through production.

The outcome

Manual statement workflows dropped by roughly 90%, and the platform became a core analytics product. I led development across the stack — from architecture and ticket breakdown to production deployment.