shubhand at cs dot cmu dot edu


नमस्ते (Namaste)! I'm a graduate student in The Robotics Institute, at Carnegie Mellon University's School of Computer Science. I'm currently being advised by Jeff Schneider on problems around designing deep reinforcement learning algorithms for safe and sample efficient training of autonomous vehicles.
Broadly, I enjoy building systems and platforms that solve challenging problems and add meaningful societal value. In the past, I've had the incredible fortune to have worked with and been mentored by:
David Thompson, with whom I designed machine learning algorithms to understand the impact of climate change on coral reefs, through my work as a Caltech Summer Research Fellow at NASA’s Jet Propulsion Laboratory.
Yogesh Wadadekar and C.H. Ishwara Chandra, under whose mentorship at the National Center for Radio Astrophysics (TIFR), an amazing team and I built distributed imaging synthesis pipelines to crunch through data collected by world's largest sub-GHz frequency radio telescope, the GMRT, and in the process enable a deeper understanding of our cosmos.
Venkat Padmanabhan and Akshay Uttama Nambi at Microsoft Research, with whom I designed streaming algorithms to efficiently analyze and determine statistical properties of data flowing through large scale networks.
I grew up in the San Francisco Bay Area before moving to India where I attended school for close to a decade. I still enjoy spending time in both countries. In my spare time I love to read books, travel, and pursue diverse forms of rapid bipedal locomotion.



ACDNN : Atmospheric Correction using Deep Neural Networks
Collaborators: David Thompson, Brian Bue, Vijay Natraj, Mario Parente
According to a recent investigation, an estimated 33-50% of the world's coral reefs have undergone degradation, believed to be as a result of climate change. However, the data supporting the investigation are scattered, and the exact relation it has to the environmental condition cannot be easily established. Through our research we designed algorithms to efficiently emulate atmospheric models using state of the art deep learning techniques. Accepted to ICML 2019 (Workshop on Climate Change + AI).
GARUDA : GMRT Archival Utility for Data Analysis
We designed a high performance compute cluster to run a data science pipeline to synthesize images from data generated by the GMRT radio telescope. Our work reduced the time for synthesis from 6+ weeks to 10-12 hours leading to the creation of one of the world's largest catalogs of radio astronomy images. Presented at the International Astronomical Union in Geneva, 36th Annual Meeting of the Astronomical Society of India, and ADASS XXXVIII (lightning talk). Published in Publications of the Astronomical Society of the Pacific.
SpiNet: Analyzing statistical properties of data in large scale datacenters
Designed and implemented novel streaming algorithms to efficiently analyze statistical properties of data in large scale datacenters. Promising results of our research have lead to an expanded inquiry and future tech transfer to Microsoft Azure.
India Digital Heritage Project : Designing an immersive VR Walkthrough
Worked on optimizing a graphics rendering pipeline for the Gear VR, which was subsequently deployed on a mobile device. Our final build had a better performance (in FPS) compared to a model designed by a similar Samsung prototype. We presented and explained our final model to Secretary, Department of Science & Technology and Secretary, Ministry of Culture, Government of India.


Sparse Approximation for Signal Recovery
For Math Fundamentals for Robotics (16-811), carried out a study of sparse approximation algorithms applied to real world problems (specifically for signal recovery in the context of astronomy).
360 Panorama Generation using Cylindrical Projections
For Computer Vision (16-720B), implemented an algorithm that generates a 360 panorama by stitching together a series of images using cylindrical projections.


GullyNet: Our Time Will Come
Presented at SIGBOVIK 2019.


Licensee, Curator, and Co-Founder
My co-founder and I started TEDxPICT with a vision to provide a platform for individuals who sought to expand their worldview through debate and the free exchange of ideas.
Though we'd started out as an informal club who'd meet to watch and discuss interesting TED talks, down the line, we streamlined our efforts through applying for a TEDx license and launching a multi-department initiative to bring light to local voices.
We built a team of over 40 enthusiastic people through several rounds of comprehensive interviews. Around this time, our core team was invited to first-ever TEDx Workshop in Mysore organized by TED HQ and Infosys, in which we were fortunate to get the opportunity to interact with fellow TEDxers from all over India.
Largely due to the hard work and efforts of our amazing team, our event went on to be a great success and was filled with intellectually stimulating talks ranging from designing student CubeSats for space satellite launches to creating devices that help people with hearing disabilities dance in sync with music.
Briefly collaborated with researchers at the Center for Policy Research, New Delhi on drawing insights from demographic data collected under the Right to Education (RTE), Act in Andhra Pradesh. The broader problem was a nice blend of optimization and statistics, and concerned balancing the tradeoff of constructing government schools closer or farther away from villages which would impact factors including the ability to maintain full staff, overall levels of student attendance, and the subsequent quality of education.


• I maintain a blog where I enjoy occasionally posting musings.

• Books are mind-expanding. They enable us to connect with and converse with wisdom from times past. I've started to maintain a selected list of influential books I've read and a short review of their content here (work in progress). You can find a more complete list here.

• I believe quotes are concise, powerful, and can profoundly influence human thought well beyond their time. You can find a sampling of the ones that I ponder on from time to time here.