Karthik Rajkumar
Staff Applied Scientist at Glean
Palo Alto, California, United States
Summary
👤
Senior🎓
Top SchoolKarthik Rajkumar is a Staff Applied Scientist blending nine years of hands-on ML engineering and econometric research to build trustworthy, evaluable AI at Glean. He designs end-to-end evaluation and experimentation infrastructure for search, assistants, and agents, and previously created core marketplace health metrics and automated conversation evaluation for LinkedIn’s first AI agent. A Stanford econometrics PhD advised by Guido Imbens and Susan Athey, he first-authored a Science paper resolving a long-standing question about weak ties using causal big data and has translated causal theory into production gains (e.g., debiasing job-ranking recall by 6 points). He’s an active contributor to GPyTorch—adding advanced Gaussian process priors and multitask likelihoods—bringing rigorous probabilistic modeling into product teams. Known for combining causal thinking, principled evaluation, and software craftsmanship, he also led LinkedIn’s Causal Inference Reading Group to upskill practitioners across academia and industry. Based in Palo Alto, he thrives at the intersection of causal inference, scalable ML systems, and reliable AI productization.
9 years of coding experience
9 years of employment as a software developer
Indian Institute of Technology MadrasDoctor of Philosophy - Ph.D. Economics Statistics, Doctor of Philosophy - Ph.D. Economics Statistics at Stanford University
English, French, Tamil, Telugu, Hindi
Github Skills (7)
pytorch10
machine-learning10
python10
gaussian-processes10
linear-algebra9
kernel-methods9
gpu-acceleration7
Programming languages (3)
Github contributions (5)
cornellius-gp/gpytorch
Aug 2018 - Sep 2018
A highly efficient implementation of Gaussian Processes in PyTorch
Role in this project:
ML Engineer
ML EngineerContributions:17 commits, 3 PRs, 12 comments in 1 month
Contributions summary:Karthik significantly contributed to the implementation and extension of Gaussian Process (GP) functionality within the GPyTorch library. Their work involved adding and modifying priors for covariance matrices, specifically focusing on LKJ priors and their application. Furthermore, the user generalized the `MultitaskGaussianLikelihood` to accommodate off-diagonal covariance structures and a diagonal matrix, demonstrating a solid understanding of GP modeling and its practical application. This included writing unit tests to validate the new features.
pytorchgpu-accelerationgaussianstochastic-processesmodular
rajkumarkarthik/mgrf-develop
Jul 2018 - Nov 2019
Working version of multivariate generalized random forests
Contributions:7 commits, 6 pushes, 1 branch in 1 year 4 months
forestsmachine-learningrandom-forestsmultivariate
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.