Sep 2018 – Mar 2020 · 1 yr 6 mos · Pasadena, California · On-site

• Responsible for building new deep learning algorithms for 2D/3D human pose estimation to create a pipeline for ObEN’s Personal AI avatars and advancing their capabilities for AI Dancing. This involved implementing a technique using domain adaptation and adversarial training to make it more robust towards in the wild data.
• Developed a novel deep learning architecture for doing Pose guided Image Synthesis using GAN’s in order to achieve 2D dance transfer given a source subject and target motion using style transfer techniques. This project also involves collecting in-house dance/action videos, building a complete data pre-preprocessing pipeline from scratch.
• Developing a new end to end pipeline for animating ObEN's virtual avatars by predicting conversational style gestures for the given subject data from raw audio/speech. This includes collecting in-house data using Motion Capture , building the entire data pre-processing pipeline and coming up with a novel deep neural network architecture for training and evaluating this audio-visual task.

Jan 2018 – May 2018 · 4 mos · Greater Pittsburgh Region

• Course: Introduction to Machine Learning(10-601), Spring 2018
• Department: Machine Learning Department
• Advisor: Matt Gormley
• Responsibilities: Designing and testing new questions, recitations, grading and holding office hours specifically for homeworks on Neural Networks and Deep learning and DImensionality Reduction.

May 2017 – Jul 2018 · 1 yr 2 mos · Greater Pittsburgh Region

• Conducted research in Deep learning and 3D Vision.
• Worked on Learning 3D volumetric cubic representations of objects from a single view 2D image in Prof. Abhinav Gupta's Computer Vision Group, Robotics Institute.
• See under Projects: "Learning 3D Volumetric Representations from 2D Images of objects" for more details.

Jun 2014 – Dec 2016 · 2 yrs 6 mos · Noida, Uttar Pradesh, India

• Design Engineer in Digital Networking System on Chip(SoC) Team working on ARM-based multicore architecture products.
• Handled design verification for timely delivery of test suites and plans for ARM A72/A53 Processor, Memory Management Unit SMMU-500 and Interrupt Controller GIC-400 for ARM v8 architecture based SoC's.
• Integrated verification environments for AMBA protocols like AXI, ACE, and CHI for these IPs.
• Conducted and managed verification of Core and L1 cache system on Freescale's StarCore (SC3900) DSP.
• Handled complete ownership of verification of Freescale's Trust Architecture building new flows to verify Low Power isolated design along the way.
• Mentored new grads on technologies and standard chip protocols in the industry.

Jan 2014 – Jun 2014 · 5 mos · Noida, Uttar Pradesh, India

• Built an Electronic Design Automation (EDA) tool for Dynamic Power optimization by modifying the actual SoC design/netlist.
• Implemented a novel way to verify Pipeline Flops in a SoC which vastly helped reduce SoC verification cycle time.

Jun 2013 – Aug 2013 · 2 mos · Bangalore

• Developed firmware for Comfort-Point Controllers for Building and Automation Control Networking (BACnet) Stack at Honeywell Building Solutions team and architected a new form of System Level Alarm detection and reporting.

Apr 2012 – Aug 2013 · 1 yr 4 mos · Udupi, Karnataka, India

• One of the Team Leads of RoboManipal Lunabotics. Participated in NASA Lunabotics Mining Competition and stood at the 19th rank from 50 teams for the Joe Kosmo Award for Excellence.
• Responsibilities: Design and development of control system for the Lunabot, systems engineering analysis and administrative responsibilities

May 2011 – Jul 2011 · 2 mos · Noida

• Image Processing in MATLAB applied in Autonomous Robotics.