Varun Budati

I've worked on projects involving order execution and optimization in crypto data, currently researching how financial LLMs manage context, benchmarking retrieval strategies like RAG and MCP. I'm also building a Python library for non-linearity testing.

I placed 15th in the US out of 12,600 teams in IMC Prosperity 3, designing market-making and signal-driven strategies across multi-asset markets. I also competed in the Undergraduate Commodities Competition, building quantitative models to price a natural gas basis trade using Monte Carlo simulations and weather derivatives.

My main stack is Python : NumPy, Pandas, PyTorch, Statsmodels, Sklearn. For web and visualization I use React, Streamlit, and Mapbox. I work with SQL and have built full-stack apps with Node.js, Flask, and FastAPI. Most of my quant work lives in Jupyter before it gets cleaned up into something shareable.

At the REACH Lab, I worked on an NSF-funded review of rural CS education. During my MAOP internship, I analyzed CS curricula from schools like MIT, UC Berkeley and Caltech and found that 57% of examples were abstract and math-heavy, leaving a lot of students behind.

Co-authored "Bridging the Code Gap: Understanding Pedagogical Patterns in CS1 Programming Courses" — a study on pedagogical patterns in CS1 courses, published in the ACM Digital Library and presented as a poster at SIGCSE 2026 in St. Louis, Missouri.

I serve as Treasurer of Virginia Tech's FinTech Club, managing our budget and operations for 100+ members. I help organize speaker events with Industry professionals, and led a faculty-advised project replicating the Evans & Archer (1968) paper on portfolio diversification.

I've mentored undergraduate researchers in literature review, database search, and research methodology, both at the REACH Lab and informally within my own networks. Getting people started on research is something I find genuinely rewarding.