-
Experience
Full Stack Developer Intern, Bombardier Transportation India Pvt. Ltd., India May 2019 – July 2019
• Designed and developed a self-service web application for company employees using PHP. Reduced company API calls by 12% by implementing caching mechanism. SMTP APIs used to send email notifications. Renovated interactive interface in JavaScript to display daily, weekly, monthly records.
Software Developer Intern, Samsung Research Institute - Noida, India January 2019 – May 2019
• Designed a Python framework for capturing various inputs of an audience namely facial expressions, emotions from voice, audience interaction, and eyeball tracking.
• Programmed to preprocess data, followed by augmentation, and extraction of emotions from a live video feed.
• Fabricated the existing Convolutional Neural Network architecture to boost model training performance that improved the accuracy of existing architecture by nearly 6% to a total accuracy of 88%.
• Integrated and tested eyeball tracking algorithm with drowsiness detection algorithm.
Software Developer Intern, Larsen & Toubro Technology Services, Vadodara, India May 2018 – July 2018
• Incorporated text auto correction module in Python to correct wrongly interpreted data from OCR. Fabricated satellite image classification module with Tensorflow backend to preprocess and classify satellite oil rig images.
-
Projects
Web-based Graphical Password Authentication System
• Constructed Java framework to capture graphical passwords. Selection of password made possible with JavaScript.
• Stored encrypted passwords on MySQL that helped prevent brute-force attacks. Hosted locally on Apache Tomcat.
Online Library Reservation System using Ruby on Rails
• Learned Ruby on Rails to develop an advanced library reservation system. RSpec used for unit tests.
• Modeled interactive fronted dashboard for users, also hosted on Heroku for online access.
AdWords Placement Problem via Online Bipartite Graph Matching
• Implemented 3 algorithms (Greedy, MSVV, & Balance) with the aim of maximizing the amount of money received from advertisers given individual budget, Adwords bids and advertisement slots.
Spatial Predictions using Satellite Images
• Tailored U-Net architecture to fit spatial satellite images. Introduced weight maps to calculate weighted loss that helped to improve results by 7.5%.