AI & AutomationLLM compliance scoring / insurance call centers
IntelligenceX
Call recordings in, LLM-scored compliance scorecards out.
- Role
- Full-stack engineer at Igknight Tech
Overview
A multi-tenant platform for insurance call centers. Recordings arrive from SFTP, S3 and partner APIs, get transcribed, then scored by configurable LLM scorecards for required disclaimers and agent performance. I worked across the ingestion pipeline, the background job stack and the React front end.
What I built
- Shipped features and production fixes across ingestion from SFTP, S3 and partner APIs
- Hardened concurrent ingestion with race-safe writes backed by unique indexes
- Added per-item isolation to scheduled job loops, so one bad record cannot stall the rest
- Improved the prompt runner behind the LLM scorecards
- Tightened the protocol for destructive production scripts, with read-only pre-flights that print the blast radius
Stack
- Ruby on Rails
- React
- PostgreSQL
- LLM APIs
- Solid Queue
- TypeScript
- Deepgram
- AWS