At Fujitsu, I designed and built an enterprise AI platform that validates Japanese-language inspection spreadsheets, combining deterministic checks with LLM-powered evaluation. Its RAG layer includes graceful fallbacks so the system can return a usable answer even when retrieval is unavailable.
I engineered the platform’s multi-phase evaluation pipeline and secure processing for MIP-encrypted Excel files, preserving VBA macros in a compliance-controlled environment. I also added incremental evaluation to avoid redundant model calls and packaged the application as a standalone Windows executable.
For Fujitsu’s Operational Efficiency Tool, I built Python REST APIs over FAISS and optimized vectorization, reducing processing time by about 39%. I also delivered Asana automation tools and ServiceNow workflows that reduced manual effort and ticket response time.
At Cognizant, I built internal tools for code analysis, monitoring, and task automation, and developed dashboards for project reporting. I’ve co-authored a peer-reviewed whitepaper on cost- and reliability-aware enterprise LLM routing, selected for Fujitsu Convention 2026, and received Fujitsu awards for AI and automation delivery.
