Reyna Abhyankar

Reyna Abhyankar Personal Page

I build efficient systems for large language models and agentic workflows. I am passionate about democratizing access to AI, open-source, and making the most out of our hardware resources.

Affiliations:

• PhD student at UC San Diego
• Systems Researcher at Together.AI (GPU Kernels)
• T5 contributor to FlexFlow-Train

Outside of work, I love playing piano and guitar, watching the Padres, and playing football.

Relevant Publications

• OSWorld-Human: Benchmarking the Efficiency of Computer-Use Agents [ICML ’25 Workshop on CUA]
• Cognify: Supercharging Gen-AI Workflows with Hierarchical Autotuning [KDD ‘25]
• Preble: Efficient Distributed Prompt Scheduling for LLM Serving [ICLR ‘25]
• InferCept: Efficient Intercept Support for Augmented Large Language Model Inference [ICML ’24]

Links

• Google Scholar
• LinkedIn
• GitHub

Contact Info

Email me at vabhyank@ucsd.edu or rabhyankar@together.ai. I use she/her pronouns.