Harivallabha Rangarajan

Software Engineer at Microsoft.

I build hyper-fast products!

About

I'm Hari, a Software Engineering graduate student at the Institute for Software Research, Carnegie Mellon University. Prior to joining CMU, I worked as a software developer at Microsoft where I was part of the Intelligence Engineering team, building copilots for Microsoft 365 Mobile. I also worked on ensuring data integrity at Exascale within the OneDrive and SharePoint organization. Currently, I work on understanding reasoning capabilities of large language models. During my undergraduate, I double majored in Mathematics and Computer Science, and minored in Data Science at the beautiful BITS Pilani - Hyderabad Campus. In my previous life, I was a national-level chess player under the tutelage of Grandmaster and former Indian national chess coach, R.B. Ramesh.

This blog serves as a compendium of thoughts - on life, love, and coding. It's primarily inteded for personal use, but is published here, in the vain hope that it is of use to someone out there as well :)

This space will host my thoughts on work, and my evolving musings on what it means to live a good life (thank you, Randy Pausch). Occassionally, I also hope to season the blog with technical topics that I find to be stimulating.

Welcome, to my pensieve! :")

Interests

Bayesian Statistics
Machine Learning
Software Engineering
Cloud Computing

Education

Master of Software Engineering, 2025

Carnegie Mellon University, School of Computer Science
M.Sc. Mathematics + B.E. Computer Science (Double Major) with a Minor in Data science, 2021

BITS Pilani - Hyderabad Campus

Recent Experience

Software Engineer

Microsoft

Jul 2021 – Present Hyderabad, India

OneDrive and SharePoint team. Developing the Fluid framework [+= Microsoft Loop] to enable seamless, real-time collaboration.

Research Affiliate

Lawrence Berkeley National Lab

Jan 2021 – July 2021 Berkeley, California

Bayesian temporal modelling for high-dimensional time series data. Developed a flexible Three-Parameter-Beta-Normal (TPBN) based global-local shrinkage prior for sparse vector autoregressions, improving over current state-of-the-art.

Remote Research Intern

Mila - Quebec AI Institute

Jan 2021 – Mar 2021 Montreal, Canada

Sample-efficient reinforcement learning from high dimensional pixel inputs on procedurally generated environments. Implemented CTNs (color transformation networks) and STNs (spatial transformer networks), and a cross-domain, cross-task transfer learning framework to achieve improved sample efficiency and generalisation.

Remote Research Intern

King's College London

Sep 2020 – Jan 2021 London, UK

Adversarial Deep Learning for Network Intrusion Detection Systems | Systems Security Lab.

Research Intern

BITS Pilani - Hyderabad Campus

Jul 2020 – Nov 2020 Hyderabad, India

GPU Accelerated Least Squares Kinetic Upwind Method Based Meshfree Solver for Aerodynamic Shape Optimization. Designed and Developed the primal version of a Meshfree Solver for Inviscid Compressible Flows, in C++, and GPU parallelized the solver with CUDA.

Software Engineer Intern

Microsoft

May 2020 – July 2020 Hyderabad, India

Computer Vision and Image Segmentation for Resource-Constrained Devices.

Remote Research Engineer

Cerenaut

Jan 2020 – May 2020 Melbourne, Australia

Moonshot Machine Learning Research and Products | Causal Models, and Explainability.

Research Intern

Monash University

Dec 2019 – Feb 2020 Melbourne, Australia

Self-Organizing Neural Network Hierarchy | [Published at AJCAI, 2020 | NAISys, 2020] Neuroscience inspired self-organizing neural network hierarchies and meta-learning.

Remote Research Intern

Technische Universität Kaiserslautern

Apr 2019 – Oct 2019 Kaisersalutern, Germany

Investigation of Spectral Regularizers for Non-Parametric Regressors. Machine Learning in Scientific Computing.

Data Science Intern

Indira Gandhi Centre for Atomic Research

Mar 2018 – Jul 2018 Kalpakkam, India

Design and Development of a Robust and Scalable Journal Recommender System, based on approximated Singular Value Decomposition and latent factor models. Implemented and deployed a system recommending journals for IGCAR Scientists, based on their publication and web-history datasets, in Python.