Evaluated and optimized large language model outputs for code generation across multiple programming languages

including Python, C++ and Typescript, improving accuracy by 15%

Developed evaluation frameworks and benchmarks using Python to assess AI model performance and training data

quality across 500+ test cases

Built RAG system at system level for log dump analysis, increasing task completion speeds by 30%

• Developed a social platform with an integrated store utilizing Next.js and Supabase for authentication and storage.

• Engineered AI agents managed by a Flask control service for real-time interactions and content moderation.

• Created a rogue MCP tool layer enabling LLM to interact with site elements while ensuring security through permissions and audit logs.

Developing a regional index to quantify Economic Policy Uncertainty in the US through web scraping, NLP, and data analysis.