|
Sachit Kuhar Hi there! I'm an AI researcher at AWS AI Labs, where I work on large language models for code and agentic software development. At AWS, I have been a core-team researcher for CodeWhisperer, Amazon’s first CodeLLM trained from scratch; the lead researcher for the Amazon Q Developer /test agent, which generates unit tests automatically; and a founding researcher for Kiro, an agentic AI-powered IDE. I’ve been fortunate to be advised by brilliant mentors, past and present, including Prof. Baishakhi Ray (AWS/Columbia), Prof. Danfei Xu (NVIDIA/GaTech), Prof. Christopher Fletcher (UC Berkeley), and Prof. Tushar Krishna (GaTech). I’m grateful for their advice and support over the years, which have helped me develop perspective across the AI stack. Before that, I received my bachelor’s degree from the Indian Institute of Technology (IIT) Guwahati. |
|
|
SpecAgent: A Speculative Retrieval and Forecasting Agent for Code Completion
[PDF]
George Ma, Anurag Koul, Qi Chen, Yawen Wu, Sachit Kuhar, Yu Yu, Aritra Sengupta, Varun Kumar, Murali Krishna Ramanathan ACL (Main Conference) 2026 Introduces a speculative retrieval and forecasting framework for improving code completion by anticipating future API usage and dynamically retrieving relevant context. BibTeX
@inproceedings{anonymous2026specagent,
title={SpecAgent: A Speculative Retrieval and Forecasting Agent for Code Completion},
author={George Ma and Anurag Koul and Qi Chen and Yawen Wu and Sachit Kuhar and Yu Yu and Aritra Sengupta and Varun Kumar and Murali Krishna Ramanathan},
booktitle={The 64th Annual Meeting of the Association for Computational Linguistics},
year={2026},
url={https://openreview.net/forum?id=sSA2VqcyCd}
}
|
|
LibEvolutionEval: A Benchmark and Study for Version-Specific Code Generation
[PDF]
[Website]
Sachit Kuhar, Wasi Uddin Ahmad, Zijian Wang, Nihal Jain, Haifeng Qian, Baishakhi Ray, Murali Krishna Ramanathan, Xiaofei Ma, Anoop Deoras Oral Presentation at NAACL (Main Conference) 2025 A benchmark analyzing code generation when libraries evolve across versions, highlighting challenges for LLM-based code completions. BibTeX
@article{kuhar2024libevolutioneval,
title={LibEvolutionEval: A Benchmark and Study for Version-Specific Code Generation},
author={Kuhar, Sachit and Ahmad, Wasi Uddin and Wang, Zijian and Jain, Nihal and Qian, Haifeng and Ray, Baishakhi and Ramanathan, Murali Krishna and Ma, Xiaofei and Deoras, Anoop},
journal={arXiv preprint arXiv:2412.04478},
year={2024}
}
|
|
UTFix: Change Aware Unit Test Repairing using LLM
[PDF]
Shanto Rahman, Sachit Kuhar, Berk Cirisci, Pranav Garg, Shiqi Wang, Xiaofei Ma, Anoop Deoras, Baishakhi Ray Oral Presentation at Object-oriented Programming, Systems, Languages, and Applications (OOPSLA) 2025 A method that leverages LLMs to detect and repair unit tests in response to code changes, enhancing software robustness. BibTeX
@inproceedings{rahman2025utfix,
title={UTFix: Change Aware Unit Test Repairing using LLM},
author={Rahman, Shanto and Kuhar, Sachit and Cirisci, Berk and Garg, Pranav and Wang, Shiqi and Ma, Xiaofei and Deoras and Ray, Baishakhi},
booktitle={Proceedings of the OOPSLA},
year={2025},
address={Singapore}
}
|
|
PLUM: Improving Efficiency By Leveraging Repetition-Sparsity Trade-Off
[PDF]
[Website]
Sachit Kuhar, Yash Jain, Alexey Tumanov Transactions on Machine Learning Research (TMLR) 2024 Spotlight Talk at MLSys On-Device Intelligence 2023 Proposes a method to exploit weight repetition and structural sparsity in neural networks to achieve better efficiency. BibTeX
@article{kuhar2024plum,
title={{PLUM}: Improving Inference Efficiency By Leveraging Repetition-Sparsity Trade-Off},
author={Kuhar, Sachit and Jain, Yash and Tumanov, Alexey},
journal={Transactions on Machine Learning Research},
year={2024},
url={https://openreview.net/forum?id=IEKtMMSblm}
}
|
|
Learning to Discern: Imitating Heterogeneous Human Demonstrations
[PDF]
Sachit Kuhar, Shuo Cheng, Shivang Chopra, Matthew Bronars, Danfei Xu Conference on Robot Learning (CoRL) 2023 Presents a method to handle mixed-quality offline demonstrations for imitation learning, improving policy performance. BibTeX
@inproceedings{kuhar2023learning,
title={Learning to Discern: Imitating Heterogeneous Human Demonstrations with Preference and Representation Learning},
author={Kuhar, Sachit and Cheng, Shuo and Chopra, Shivang and Bronars, Matthew and Xu, Danfei},
booktitle={7th Annual Conference on Robot Learning (CoRL)},
year={2023},
url={https://openreview.net/forum?id=kOm3jWX8YN}
}
|
|
Offline Visual Representation Learning for Embodied Navigation
[PDF]
Karmesh, Ram, Arjun, Vincent, Sachit Kuhar, Dhruv Batra, Alexei Baevski, Oleksandr Maksymets ICLR Reincarnating Reinforcement Learning 2023 Examines self-supervised learning approaches for visual encoders in embodied navigation tasks using offline datasets. BibTeX
@inproceedings{yadav2023offline,
title={Offline Visual Representation Learning for Embodied Navigation},
author={Karmesh Yadav and Ram Ramrakhya and Arjun Majumdar and Vincent-Pierre Berges and Sachit Kuhar and Dhruv Batra and Alexei Baevski and Oleksandr Maksymets},
booktitle={Workshop on Reincarnating Reinforcement Learning at ICLR 2023},
year={2023},
url={https://openreview.net/forum?id=Spfbts_vNY}
}
|
|
SumMerge: Algorithm and Implementation for Weight Repetition-Aware DNN Inference
[PDF]
Rohan Prabhakar*, Sachit Kuhar*, Rohit Agrawal, Christopher Hughes, Christopher Fletcher Oral Presentation at International Conference on Supercomputing (ICS) 2021 Introduces an algorithm to accelerate DNN inference by exploiting weight repetition patterns, showing 2x improvement on Intel CPUs. BibTeX
@inproceedings{prabhakar2021summerge,
title={Summerge: An efficient algorithm and implementation for weight repetition-aware dnn inference},
author={Prabhakar, Rohan Baskar and Kuhar, Sachit and Agrawal, Rohit and Hughes, Christopher J and Fletcher, Christopher W},
booktitle={Proceedings of the ACM International Conference on Supercomputing},
pages={279--290},
year={2021}
}
|
|
mRNA: Enabling Efficient Mapping Space Exploration for a Reconfiguration Neural Accelerator
[PDF]
Zhongyuan Zhao, Hyoukjun Kwon, Sachit Kuhar, Weiguang Sheng, Z Mao, Tushar Krishna Oral Presentation at International Symposium on Performance Analysis of Systems and Software (ISPASS) 2019 Proposes a design-space exploration methodology for mapping DNNs efficiently on a reconfigurable neural accelerator. BibTeX
@inproceedings{zhao2019mrna,
title={mrna: Enabling efficient mapping space exploration for a reconfiguration neural accelerator},
author={Zhao, Zhongyuan and Kwon, Hyoukjun and Kuhar, Sachit and Sheng, Weiguang and Mao, Zhigang and Krishna, Tushar},
booktitle={2019 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS)},
pages={282--292},
year={2019},
organization={IEEE}
}
|
|
Deep Learning based Semi-Blind Tracking for Aging Wireless Communication Channels
[PDF]
Sachit Kuhar*, Achal Dave*, Ribhu Chopra Springer Wireless Personal Communications (WPC), 2021 Develops a novel neural approach to track changes in wireless channels, improving communication reliability over time. BibTeX
@article{dave2021deep,
title={Deep learning based semi-blind tracking for aging wireless communication channels},
author={Dave, Achal and Kuhar, Sachit and Chopra, Ribhu},
journal={Wireless Personal Communications},
volume={119},
number={3},
pages={2695--2706},
year={2021},
publisher={Springer}
}
|