Sreeshanth Ryali

• Designed a multi-agent financial reasoning system to generate personalized financial recommendations
• Built a restricted Python execution tool that lets the LLM write and run custom financial computation code, reducing arithmetic errors
• Implemented a RAG system with a global knowledge base and per-user vector stores, integrated with real-time market data via Tavily and yfinance

Python, FastAPI, LangGraph, FAISS

• Engineered a tool-using AI agent, enabling natural-language control over web search, Google Calendar, and Gmail workflows
• Designed a dual-LLM verification pipeline (generator + reviewer), reducing hallucinations through iterative refinement
• Implemented an observability interface exposing execution traces and LLM-to-LLM critiques for transparent agent debugging

Python, FastAPI, LangChain, LangGraph

• Full-stack Kanban board for professors with drag-and-drop task management and session-based auth
• Automated course planning by integrating LLMs to parse PDF syllabi and generate structured tasks
• Custom probability distribution from real exam patterns, generated ~9,000 synthetic records
• Trained a classifier to predict final grades from mid-term scores with noise injection

Python, Flask, MySQL, scikit-learn, LangChain