Applied AI
Fuel iX
Agent Quality Management (AQM)
Backend architecture and delivery for LLM contact-center conversation evaluation.
As Backend Tech Lead on Agent Quality Management (AQM), I joined during alpha to drive the backend architecture and ship the product to production. I re-architected the conversation ingestion pipeline into a live API gateway, led technical alignment across three engineering teams, and hardened system security and operational stability for the December 2025 production launch.
CONTEXT
Project overview
Agent Quality Management (AQM) is a flagship Fuel iX product that applies Large Language Models to customer contact-center conversations to automatically evaluate agent performance and ensure quality standards at enterprise scale. Transforming the platform from prototype into an enterprise solution required replacing mock pipelines with live conversation ingestion, secure identity mapping, and robust API boundaries.
CONTRIBUTION
My contribution
Joined during the early alpha phase as Backend Tech Lead to steer backend architecture, coordinate cross-team integration across AQM, VCon, and Fuel iX core, and drive the product through its November alpha delivery and December 2025 production launch.
Live transcript ingestion & API gateway architecture
Owned the critical integration between AQM and the VCon transcript pipeline, bridging mock data to real-world customer conversations. Designed and implemented the identity mapping API between systems, replaced mocked agent identifiers with live endpoint resolution, and re-architected the alpha ingestion flow into a scalable, decoupled API gateway.
Cross-team alignment & technical governance
Led technical alignment across three distinct engineering teams—AQM, VCon, and Fuel iX core—collaborating with product management and engineering leadership to define role-based access requirements, service contracts, and feature-flag rollout strategies.
Production hardening & security modernization
Hardened the service for enterprise production by implementing database connection pooling, container startup and health probes, least-privilege service accounts, and transitioning authentication from legacy static API keys to secure Firebase ID tokens.
Release leadership & engineering standards
Directed release operations through Fuel iX engineering governance channels, resolved CI secrets and infrastructure deployment blockers, conducted pull request reviews, resolved type errors in shared codebases, and established clear architectural boundaries between frontend and backend responsibilities.
IMPLEMENTATION
Architecture & stack
NestJS microservices containerized and hosted on Google Cloud Run, integrating with VCon transcript streams and LLM evaluation engines. Backed by PostgreSQL with connection pooling, authenticated via Firebase ID tokens, and monitored with automated startup probes and structured logging.
- NestJS
- TypeScript
- Google Cloud Run
- LLMs
- vCon
- PostgreSQL
- Firebase
- Docker
- GCP
DELIVERY
What I delivered
Shipped the AQM alpha in late November 2025 and launched the first production release on December 10, 2025.
Scalable API gateway and live agent identity mapping service connecting AQM with the VCon transcript pipeline.
Hardened enterprise security with Firebase ID token authentication, least-privilege service accounts, and database connection pooling.
Cross-team feature-flag gating and service contracts aligned across AQM, VCon, and Fuel iX core.