Lahoti Earns Jump Trading Fellowship
Adam KohlhaasFriday, September 12, 2025Print this page.
Aakash Lahoti, a Ph.D. student in Carnegie Mellon University's School of Computer Science, has been named a 2025-2026 Jump Trading Fellow for his work on building efficient, effective sequence-to-sequence models that can handle very long inputs.
Designed to support high-impact research in artificial intelligence, machine learning and other STEM fields, the Jump Trading Fellowship provides a year of support, including tuition, fees and a stipend of about $50,000. Lahoti joins two other recipients from disciplines including astrophysics and probability theory.
Lahoti, who is part of CMU's Machine Learning Department, studies the limits of attention-based architectures, which become prohibitively expensive as input length grows. His current focus is on pushing the performance frontier of fast, linear time models under tight compute budgets — improving model capabilities while reducing inference time and memory. His interests also include extending these models to structured domains like graphs and exploring applications in areas such as scientific computing, where long-range dependencies and scale often remain beyond the reach of traditional approaches.
"We're grateful to Jump for this support. This fellowship helps accelerate our push to close and surpass the quality gap with attention models while lowering latency and memory cost for broader, cost-efficient deployment," Lahoti said.
For more information, visit the Jump Trading fellowship program website.
Aaron Aupperlee | 412-268-9068 | aaupperlee@cmu.edu