Ashburn, Virginia, United States
Hey! I'm a 2ⁿᵈ year at MIT majoring in 6-3 & 18 (CSE + math). I love solving hard problems, particularly in the areas of decentralized AI, multimodal fusion, and reinforcement learning. I'm currently doing research at the MIT Laboratory for Information and Decision-Making (LIDS), optimizing strategies for dynamic compute allocation at inference. I spend a lot of time outside class working on various side projects, both creative and technical, and I live for meaningful conversations. For fun, I like finding new running routes, sampling caramel macchiatos from local cafes, and writing stories/prose/poetry. Feel free to reach out, I absolutely adore meeting new people. :)
Inference architecture 🔨 (we're hiring!)
Investing in Boston's most ambitious student founders.
On the AWS DocDB team, I utilized native query protocol and built on Aurora Postgres (APG) 17 to support the functionality of new operators, constructing new date and time series operators with enhanced parsing and evaluation processes for queries. I also researched scalable, low-latency inference in vLLM via prefill-decode (PD) disaggregation, leveraging LMCache for KV Caching, at the AWS AI & CV Labs. I optimized for production deployment under real-world workloads, fine-tuning chunk size, cache size, and decode strategies to improve TTFT and ITL.
My research at MITRE Labs focuses on geopolitical risk modeling and conflict prediction. I use reinforcement learning techniques and math modeling to derive optimal conflict mitigation strategies and inform public policy, constantly seeking out ways to improve accuracy and time-sensitivity in my models.
During this one-month micro internship, I analyzed records of accounting data for small and medium-sized businesses to predict debt, assess financial risk, & support loan underwriting decisions. I also implemented time-series modeling to extract seasonality from revenue figures and assess future assets of a company.