PhD Student in Computer Science at Columbia University
I am a third-year PhD student advised by Professor Suman Jana. My research advances the robustness and interpretability of code generation models. My recent work examines how these models generate code under adversarial conditions, comprehend long-context code, and exhibit distinct strengths that enable interpretable routing. My goal is to make code AI systems performant and trustworthy for real-world use.
Routesplain: Towards Faithful and Intervenable Routing for Software-related Tasks [paper]
, Vikas Upadhyay, Marianne Menglin Liu, Daniel W. Peterson, Anshul Mittal, Sujeeth Bharadwaj, Fahad Shah, Dan Roth
Sense and Sensitivity: Examining the Influence of Semantic Recall on Long Context Code Reasoning [paper]
, Mukur Gupta, Samira Hajizadeh, Prashast Srivastava, Suman Jana
XOXO: Stealthy Cross-Origin Context Poisoning Attacks against AI Coding Assistants [paper]
, Mukur Gupta, Noopur Bhatt, Aditya Gupta, Janie Kim, Prashast Srivastava, Suman Jana
Languages: Python, MATLAB, C/C++, Java, Go, Rust, x86, x64
Skills: Cloud Compute (AWS, GCP, Azure), Linux, PyTorch, transformers, vLLM, Pyro, IDA Pro, Ghidra, LLVM