Naman Jain

465 Soda Hall

Berkeley, California 94709

I am a PhD student at UC Berkeley where I work in Programming Languages and Machine Learning. I am particularly interested in how machine learning techniques can be used for developer productivity tools like program analysis, synthesis, and repair (ML for PL). I am also excited about how synthesis and verification can help in designing better generalizing and explainable algorithms and also reasoning about such algorithms (PL for ML)

I recieved my undergraduate degree from IIT Bombay where I was advised by Prof. Sunita Sarawagi on understanding robustness in natural language processing. Prior to that, I also delved quite a bit into computer vision (mostly adjacent to human-pose-estimation) with Prof. Arjun Jain and Prof. Abhinav Shrivastava at UMD College Park.

After my undergraduate, I spent two amazing years as a predoctoral research fellow at Microsoft Research India where I worked on machine learning and/or program synthesis. I was part of the Project Jigsaw and worked on cool research problems like program-repair with static analysis tools, improving large language models with guarentees, and learning decision trees with bandit feedback.

News

Dec 2021	Jigsaw (combining program synthesis with large language models) has been accepted at ICSE’22
Oct 2021	Preprint out on learning decision trees with bandit feedback
Aug 2020	Joined Microsoft Research, India as a Research Fellow
May 2020	Paper on named entity robustness of BERT models accepted to RepL4NLP, ACL’20
Jun 2019	Visiting University of Maryland for research internship with Prof. Abhinava Shrivastava
May 2019	Paper on adversarial examples in human pose estimation accepted in VUHCS, CVPR’19
Dec 2018	Recieved Undergraduate Research Award (URA) from IIT Bombay

Academic Advisors

I have been fortunate to work many great researchers

Selected Papers

Jigsaw: Large Language Models meet Program Synthesis

Naman Jain, Skanda Vaidyanath, Arun Iyer, and 4 more authors

ICSE May 2022

Bib

@article{Jain2021Jigsaw,
  title = {Jigsaw: Large Language Models meet Program Synthesis},
  author = {Jain, Naman and Vaidyanath, Skanda and Iyer, Arun and Natarajan, Nagarajan and Parthasarathy, Suresh and Rajamani, Sriram and Sharma, Rahul},
  journal = {ICSE},
  month = may,
  year = {2022},
}

Learning Accurate Decision Trees with Bandit Feedback via Quantized Gradient Descent

Ajaykrishna Karthikeyan, Naman Jain, Nagarajan Natarajan, and 1 more author

Aug 2022

Bib

@article{Karthikeyan2021Learning,
  title = {Learning Accurate Decision Trees with Bandit Feedback via Quantized Gradient Descent},
  author = {Karthikeyan, Ajaykrishna and Jain, Naman and Natarajan, Nagarajan and Jain, Prateek},
  booktitle = {TMLR},
  month = aug,
  year = {2022},
}