At KPMG, I build production LLM agents end to end, including a multi-agent platform processing 10,000+ requirements. It reduced manual QA by 65% and made releases four times faster, with CI/CD at 99.9% uptime.
I own the evaluation layer, with 43/43 eval suites green and regression gating before every release. Prompt A/B testing cut false alarms from four to zero, and I log cost and latency across live agent runs.
I also ship production RAG and matching systems using Azure AI Search and Azure OpenAI. They index 500K+ documents, deliver sub-200ms queries, improve match accuracy by 40%, and reduce screening time by 75%.
In my personal projects, I built a GitHub pull-request review agent and a RAG Q&A API. Both have passing test suites, and my portfolio includes live demos of these projects.
