Jan 2021 – Apr 2021 · 3 mos

• Reporting to CEO, I was responsible for developing AI Strategy.

Jul 2019 – Oct 2020 · 1 yr 3 mos · Pittsburgh

• Few Shot Learning methods for Object Detection in the Wild!
• Advisors: Dr. Jeff Schneider and Dr. Christoph Mertz
• Benchmarking Off-the-shelf model: Framework for benchmarking any off-the-shelf model
• against ImageNet (ILSVRC) to ensure model file integrity.
• Few Shot Transfer Learning in the wild: Developed Few-Shot Transfer learning algo-
• rithms on a custom dataset containing few hundred images which involving less than 1 hour
• of end-to-end training time for multi-class object detection.
• Rapid End-to-End Transfer Learning framework: Extended above few shot learning
• feature to include desired hyperparameter selection for the backbone network, handle multiple
• data streams and perform model selection, validation and benchmarking on test set.
• Communication: Showcased performance as a part of sponsor demos, interacted with part-
• ner institutions and incorporated feedback.

Oct 2018 – Jun 2019 · 8 mos · Washington D.C. Metro Area

• Design and develop accurate real-time object detection using Deep learning based framework on Nvidia Jetson TX2.
• Additional Responsibility: Designed and implemented solution for remote control of ground drone movement by integrating Sony Playstation 3 controller with onboard Nvidia Jetson TX2 platform. Also, developed motor control scripts for wide variation in drone movement from fine-grained to coarse depending upon requirement.

Feb 2018 – Oct 2018 · 8 mos

• - Design and develop computer vision algorithms to perform real-time pedestrian traffic light detection to aid motion of Unmanned Ground Drone.

Oct 2017 – Jan 2018 · 3 mos

• Design and develop computer vision algorithms as part of vision module to perform pedestrian light and pedestrian crossing recognition to aid motion of Unmanned Ground Drone.

Nov 2017 – Jan 2018 · 2 mos

• Explore semi-supervised topic modeling methods for analysing tweets referencing Venezuelan protests.

Jan 2016 – Oct 2016 · 9 mos · College of Behavioral and Social Sciences, University of Maryland, College Park

• Designed and developed algorithms for identifying relevant tweets referencing events of interest from Twitter Firehose using unsupervised methods.
• Performed independent study titled "Measuring grievance through text" in collaboration with Dr. Erin McGrath sentiment analysis of tweets referencing Nigerian 2015 and South African 2014 elections to identify tweets referencing hate as an indicator of grievance

Jan 2015 – May 2015 · 4 mos

Mar 2014 – Jan 2015 · 10 mos · Mandi Area, India

• Developed a state-of-art text-to-speech systems using Festival and HTS for Rajasthani, thus enabling speech interface for information access to aid visually challenged people.
• Main contribution involved improvement in synthesized voice quality for statistical parametric synthesis using pitch tracking methods.

Aug 2010 – Jan 2014 · 3 yrs 5 mos · Chennai Area, India

• Developed a state-of-art automatic speech recognition application in Tamil, a native Indian language to provide up-to-date district-wise price of queried commodity from website.
• Contributed as a part of think-tank, concept, design and implementation of the User Interface, back-end SQL database and web-crawler. Experimented with acoustic and language models to reduce word error rates (WER), thus improving recognition accuracy.