At Movate / KKR, I built tool-using LLM agents with Claude Code and custom MCP servers that connect to GitHub, Terraform, AWS, and observability APIs. They fix CI/CD failures, propose infrastructure changes, and generate runbooks, with each change routed through a pull request for human approval.
I also built a LiteLLM model layer spanning Amazon Bedrock, SageMaker, Vertex AI, and Azure AI Foundry, along with an internal RAG chatbot and production AI services. Earlier, at EPAM Systems, I built an LLM-assisted incident workflow that cut MTTR from 90 to 15 minutes across 200+ incidents, and at Luxoft I worked on in-vehicle speech-recognition AI serving 2M+ vehicles.
