Jan 2020 – Apr 2020 · 3 mos · Atlanta Metropolitan Area

• CS 6301: Advanced Software Engineering

Aug 2019 – Dec 2019 · 4 mos · San Francisco Bay Area

• Autonomous Systems, Special Projects Group

May 2019 – Aug 2019 · 3 mos · Greater Pittsburgh Region

• Perception, Uber Advanced Technologies Group (ATG)

Jan 2019 – Apr 2019 · 3 mos · Greater Atlanta Area

• Smart City Infrastructure Health and Safety Condition Assessment and Monitoring, sponsored by the US Department of Transportation (USDOT), Federal Highway Administration (FHWA), NCHRP Innovation Deserving Exploratory Analysis (IDEA) program, and Georgia DOT:
• Automatic detection and classification of traffic signs using deep learning with TensorFlow

Apr 2018 – Jul 2018 · 3 mos · Bengaluru, Karnataka, India

• Vision Research group
• Model compression:
• Translated a TensorFlow model of a neural network for image super-resolution to a fixed-point implementation in C for Samsung TVs
• Decreased inference time by 60% and memory usage by 75% due to reduced precision, reduced memory consumption by 46% via buffer reuse
• Point cloud compression:
• Devised a novel point cloud compression algorithm based on camera projection, achieved a compression ratio of 700 and PSNR 70 dB. Filed a provisional application with the Indian Patent Office.

Jun 2016 – Mar 2018 · 1 yr 9 mos · Bengaluru, Karnataka, India

• Vision Research group
• 3D reconstruction from images:
• Developed an end-to-end pipeline for 3D reconstruction comprising autonomous image capture using unmanned aerial vehicles, structure-from-motion and bundle adjustment for camera pose estimation and sparse point cloud generation, computation of depth maps and their fusion to create a surface mesh with colored texture
• Built a novel machine learning system to predict the quality of 3D reconstruction using only low-level image features with an accuracy of 94%
• Brain-Computer Interfaces:
• Implemented signal processing and machine learning algorithms for emotion recognition and motor imagery classification from EEG signals
• Explored methods for automatic channel selection for classification from a small number of EEG channels
• Designed a new CNN architecture for EEG-based, subject-independent motor imagery classification, achieved 91% accuracy (state-of-the-art)
• Filed a patent application with the Indian Patent Office on detecting and correcting virtual reality sickness using brain-computer interfaces

May 2015 – Jul 2015 · 2 mos · Bengaluru, Karnataka, India

• Linear registration of MRI images and 3D volumes using the Lucas-Kanade method: reduced computation time by 53% by developing a parallel implementation of the algorithm in an inverse compositional framework in C++.

Apr 2015 – Mar 2016 · 11 mos · Mumbai Metropolitan Region

• Insight is one of the world's most followed English student media bodies (https://www.insightiitb.org/)
• Led a 19-membered Editorial Board to bring out newsletters and web articles reaching out to more than 650 faculty members and 10,000 students on campus on a regular basis
• Collaborated with The New York Times for a global project titled "The Secret Life of Passwords"
• Conducted a comprehensive survey of the graduating batch and the freshman batch in collaboration with other IITs, covered by Indian national dailies such as The Times of India, Hindustan Times
• Achieved 30% y-o-y increase on Facebook (70k likes), 56% y-o-y increase in YouTube subscribers, 307k+ web views and 1.1 million YouTube views

Dec 2014 – Jan 2015 · 1 mo · Mumbai Metropolitan Region

• Examined various global optimization methods for calibration of stochastic financial models
• used to price interest rate derivatives
• Calibrated the SABR and the Two Factor Hull-White models using the Differential Evolution algorithm

Jul 2014 – Nov 2014 · 4 mos · Mumbai Metropolitan Region

• Designed and conducted tutorials for over 100 students and evaluated their performance.