Apr 2024 – Present
Staff Engineer — AI and Backend Architecture
- Architected a LangGraph/LangChain multi-agent RAG platform serving approximately 1,000 users at 100 requests per minute, combining BM25 and vector retrieval for enterprise product intelligence; contributed golden datasets for RAGAS evaluation.
- Built autonomous ticket triage and root-cause analysis across 87+ Jira boards, using client-specific knowledge and MCP for SLA-driven updates and remediation recommendations.
- Implemented TTL cache refresh and CDC-triggered invalidation for Customer Portal reads, bypassed caching for account balances, and enabled multi-agent tool calling over shared Lessor/Payment APIs.
- Reduced token and cloud compute costs by 35% through containerized local/cloud LLM integration while supporting on-premises enterprise data requirements.
- Used local LLMs for PII checks and request-data validation before forwarding requests to frontier models.
- Cut my individual delivery time by approximately 80% using Claude Code and Codex CLI for parallel development, debugging, agent context management, and code review.
- Handle daily client calls to gather and clarify requirements, translating business needs into backend features and AI workflows.