About me

I'm a graduate student studying Computer Vision and Machine Learning at the Robotics Institute, Carnegie Mellon University. Last summer, I interned with the Perception semantics team at Waymo, working on large Vision Language Models (VLMs) to solve the long tail edge scenarios. I'm broadly interested in problems at the heart of visual perception, image synthesis, multi-modal learning, robotics, and all things machine intelligence. I have past experience in building real-time scalable data systems to address some of these challenges.

At CMU, I am working with the Air Lab in the Robotics Institute, on multi-view stereo Depth prediction from 6-pair fisheye camera lenses for autonomous navigation. Earlier, I worked with the Xu Lab in the School of Computer Science as a research assistant, exploring visual learning pipelines for self-supervised extraction of 3D object-aware representations using controllable GANs and domain adaptation. I also TA'ed the popular Machine learning course from the ML Department over the Spring (392 students) and Fall (487 students) semesters in 2023.

Prior to grad school, I led the India data science team at AppOrchid Inc's R&D Division. My research problems here were skewed towards Document Representation learning and Semantic PDF understanding for financial doc cohorts where I extract metadata from legalese docs using DL & traditional vision techniques. My work spanned across representation learning problems in CV, resource constrained-ML, and statistical hypothesis analysis for handling long-tails in OOD inference.

I was an OSS contributor and technical mentor at OpenMined, exploring my research interests in PPML ∩ Differential privacy. I also spent a year at Soliton tech designing end-to-end real time stereo image processing pipelines to control $2MM pilot industrial systems. I graduated from Anna University, India with a Bachelor’s degree in Electrical Engineering. My past research interests were in the interdisciplinary field of electrical motor design optimization using statistical analysis tools, and I still relish contemporary research here.

P.S. This site was recently made and is a little sparse - you can find a summary of my past work here. I'm also available at sivarams@cs.cmu.edu or my socials.