I am a Ph.D. student at MIT CSAIL working with Prof. Christina Delimitrou.

Current Research

My research develops reusable mechanisms and recipes that enable safe and efficient integration of machine learning into computer systems.

• Resilient ML-driven systems: Using uncertainty as a primitive to make ML-driven systems more generalizable, interpretable, and fast under distribution shifts. [MLSys 2026, CAL 2024].
• Agents for system design and operation: Building agents that help design and operate computer systems, from hardware–software co-design to cloud performance debugging, and using uncertainty so they act when confident and escalate when not [Coco, QuArch].

Background

Before my current line of work, I spent formative years at IIT Gandhinagar working with Profs. Manu Awasthi, Joycee Mekie, and Rajeev Balasubramonian on approximate computing (Fixed-Posit), memory systems (HotStorage 2020), and mobile architecture (META, TBench 2021).

Early on at MIT, for my Master’s, I built a frequency-aware request scheduler for microservices by modifying the Thrift RPC framework. Later on, I contributed to serverless systems (Sabre @ OSDI) and carbon-aware scheduling systems (SoCC 2024).

Community

• I maintain ml-for-systems.github.io to track ML-for-Systems papers in top-tier systems venues.
• I regularly review for the ML for Systems@NeurIPS and MLArchSys workshops.
• I regularly TA for the Datacenter Computing and Computation Structures courses at MIT.