Call-Compliance AI Pipeline
Scores every agent call against compliance rules and routes violations automatically.
- Role
- Engineer
- Client
- US fintech (anonymized)
- Year
- 2026
- Lane
- AI & Data Platform
The problem
A regulated call center had no scalable way to check whether agents followed compliance scripts. Manual QA sampled a fraction of calls; violations slipped through and surfaced late.
What I built
An event-driven pipeline that ingests every call from the call-analytics provider, stages transcripts in Snowflake, scores them against configurable compliance rules, and routes violations — with an escalation engine — into Teams, Jira, and the training system. A separate Next.js config UI lets compliance staff edit rules with zero client-held credentials.
Outcome
Every call scored, not a sample — with violations escalated automatically to the right people and system, replacing manual spot-checks with continuous coverage.