AIOps Engineer
TCS (contractor at Roche) · Madrid, Spain
- Led the design and built the core of an operational knowledge graph and ontology over the company's observability, ITSM, CMDB and on-call systems. It stores no data: every query resolves live from the systems of record. Exposed to AI agents via an MCP server and a web UI, with per-user audit and role-based access. Served 5,000+ queries, 90% from AI agents; 88% answered with data at 0.6 s p50 latency.
- Built an AI incident response agent (LangGraph, Claude, RAG + ChromaDB) that diagnoses incidents by correlating monitoring episodes, tickets and operational knowledge, and graduated it from L1 (suggest) to L2 (act under guardrails).
- Built the evaluation layer for agents and skills: gold/silver/bronze datasets from human-verified incidents, weekly LLM-as-judge runs and usage telemetry. Defined autonomy levels that gate what every agent can do unsupervised.
- Own a self-service recovery automation system: ~950 automated remediations/month across 4 orchestrators, used by 10+ teams to run their own recovery workflows.
- Designed and built a PagerDuty Account Vending Machine: self-service platform (Terraform modules, Terraform Cloud, GitOps with Python + Jinja2 + GitLab CI/CD), adopted as the team's production onboarding system.
- Built 15+ AI-powered skills automating sprint planning, incident management, documentation authoring and cross-system troubleshooting.
- Run the central alert-correlation and incident-routing platform for 50+ application teams.