Divyat Mahajan
Visiting Researcher at Meta
Montreal, Quebec, Canada
Summary
🤩
Rockstar🎓
Top SchoolDivyat Mahajan is a machine learning researcher-engineer with 10 years of experience focused on robust, causally grounded AI and representation learning, currently based at Mila and a visiting researcher at Meta working on language model pretraining and compositional generalization. He holds a PhD in Computer Science from Université de Montréal and a dual BSc/BTech from IIT Kanpur, and has contributed research and internships across Microsoft Research, Aalto, and NUS on topics from amortized inference to recommender systems. His open-source work includes implementing BaseGenCF modules for the interpretml/DiCE project—bringing VAE-based counterfactual generation and PyTorch integration to a widely used explainability library. Colleagues know him for bridging rigorous theory (causal and OOD generalization) with practical implementations that improve model reliability in real-world pipelines.
10 years of coding experience
Doctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at Université de MontréalIndian Institute of Technology Kanpur
English, Hindi
Stackoverflow
161reputation
15kreached
0answers
13questions
Github Skills (19)
pytorch10
xai10
python10
model-driven10
machine-learning10
model-building10
explainable-artificial-intelligence10
deep-learning10
modeling10
model-driven-development10
data-science9
computer-vision6
android6
google-cloud-vision6
bash6
Programming languages (2)
Github contributions (5)
interpretml/DiCE
Jan 2020 - Jul 2020
Generate Diverse Counterfactual Explanations for any machine learning model.
Role in this project:
ML Engineer
ML EngineerContributions:39 commits, 3 PRs, 37 pushes in 5 months
Contributions summary:Divyat's primary contributions revolve around the implementation and refinement of the BaseGenCF method for generating counterfactual explanations. Their work involved adding modules for model approximation and oracle methods within the `dice_ml` library. The commits show modifications to the model architecture and loss functions within the context of a Variational Autoencoder (VAE) for counterfactual generation, along with integrating the model with PyTorch. Additionally, the user addressed data loading issues within a feasible counterfactuals notebook.
explainable-mlcounterfactual-explanationsdiversedataminingxai
microsoft/robustdg
Jun 2020 - Mar 2022
Toolkit for building machine learning models that generalize to unseen domains and are robust to privacy and other attacks.
Contributions:155 commits, 25 PRs, 109 pushes in 1 year 9 months
machine-buildingfairness-mlpythonunseenprivacy
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.