Dipika Khullar
Research Fellow at ML Alignment & Theory Scholars
San Diego, California, United States
Summary
🤩
Rockstar🎓
Top SchoolDipika Khullar is a research-focused machine learning engineer with five years of experience building and optimizing ML systems across industry and research labs, currently working on pretraining at Amazon and as a Research Fellow at Anthropic. She has deep hands-on expertise in model compression and quantization—contributing practical examples to the widely used AIMET toolkit—and has prototyped neural architecture and classifier-head innovations at Apple. Her work spans applied production modeling (Square, Qualcomm) to academic research in assistive video captioning at Berkeley, giving her a strong bridge between novel research and deployable solutions. Known for falling in love with problems rather than just solutions, she combines curiosity-driven investigation with pragmatic engineering to push model efficiency and alignment efforts forward.
5 years of coding experience
2 years of employment as a software developer
Bachelor of Arts - BA Computer Science, Bachelor of Arts - BA Computer Science at UC Berkeley Electrical Engineering & Computer Sciences (EECS)Minor Data Science, Minor Data Science at University of California, Berkeley
English, Hindi
Github Skills (6)
quantization10
pytorch10
machine-learning10
deeplearning-ai10
deep-learning10
python10
Programming languages (1)
Github contributions (3)
quic/aimet
Jun 2021 - Feb 2022
AIMET is a library that provides advanced quantization and compression techniques for trained neural network models.
Role in this project:
ML Engineer
ML EngineerContributions:2 reviews, 24 commits, 28 PRs in 8 months
Contributions summary:Dipika added example code for model compression using AIMET's Weight SVD technique on a ResNet18 model, demonstrating compression and finetuning workflows. They implemented compression using the auto mode, configuring target compression ratios and ignoring specific model layers. The user integrated the weight SVD compression with existing ImageNet data pipelines for evaluation and finetuning. The commits included the example file.
pytorchtechniquesdeep-learningpruningcompression
quic-dkhullar/aimet
May 2021 - Feb 2022
AIMET is a library that provides advanced quantization and compression techniques for trained neural network models.
Contributions:2 pushes, 22 branches in 9 months
pytorchtechniquesdeep-learningcompressionmachine-learning
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.