Software Developer

Kytheon

August 2019 - Present

• Developed Mood detection, Hand Raise detection, and Fake Product detection systems in Python, achieving 90%+ accuracy, and created REST APIs for the same using FastAPI and Flask.
• Implemented real-time Hand Raise and Mood detection in the browser using ReactJS and the user's webcam, reducing response time by 50%.
• Designed and developed a reusable Video/Image upload modal in ReactJS, decreasing development time for similar features by 20%.
• Contributed to the development of the CMS and frontend for an edutech product using JavaScript and ReactJS, serving 10,000+ users.
• Implemented Computer Adaptive Testing algorithms in Python using catsim, improving test efficiency by 15%.
• Developed an automated script in JavaScript for fund transfer using web3 deployed on AWS Lambda, processing 1000+ transactions daily.
• Designed and implemented tax data generation and calculation processes for a DeFi project in Python, reducing manual calculation errors by 95%.
• Developed a centralized bridge aggregator in TypeScript using web3, providing users with optimal token transfer options across multiple chains and increasing user engagement by 25%.
• Led the design, development, and maintenance of backend services and infrastructure for core blockchain-based products, ensuring 99.9% uptime.
• Developed a Flask-based application for user photo uploads and digital scrollable photo frame generation, handling 1000+ images per day.
• Designed and implemented a Natural Language-based Semantic Search engine in Python and MySQL, improving search query accuracy by 30%.
• Developed an automated duplicate photo detection algorithm using deep learning in Python, achieving 95% accuracy.
• Developed a scalable web framework using web sockets in Java, handling 10,000+ concurrent GET and POST requests.
• Designed and implemented an in-memory SQL database based on a hashmap data structure in Java, achieving O(1) query time complexity.
• Developed an e-commerce website using JSP and Servlets which included a Razorpay payment gateway integration, resulting in 200+ daily transactions.
• Developed a ticketing website for booking tickets for a cultural event using JSP and Servlets which included a Razorpay payment gateway for buying tickets.
• Developed a general-purpose recommendation engine using counters and matrices in Java.
• Collaborated with the quality assurance team to verify correct implementation of software development changes.
• Designed web-based assets for deployment on Amazon Web Services (AWS) cloud computing platforms.
• Leveraged mathematical techniques to develop engineering and scientific solutions.

Hybrid Cloud Intern / IBM India Software Labs

IBM

January 2019 - July 2019

• Developed an Angular web application to provide a single view of transaction lifecycles over the Sterling Integration Cluster, improving monitoring efficiency by 40%.
• Implemented features such as a completely editable data table, dynamic lists, and color-coded trees in Angular to enhance the user experience for B2B transaction monitoring.
• Developed a J2EE web application-based parser in Java to automate invoice processing, saving 8-15 hours of manual work per invoice.
• Worked on Face detection through the users webcam using OpenCV and deep learning in Python.
• Developed an automated system which stores the user inputed pattern or gesture through the webcam and matches it in real time with the feed from the webcam or any other camera using deep learning in Python.
• Developed a code generation utility using decision trees in Python.
• Developed reporting tools for internal usage using JAVA Swing and CouchDB.

Data Scientist Intern

RailYatri

June 2018 - July 2018

• Developed an R Shiny Dashboard for the PR Team to visualize train delays based on various metrics, improving reporting efficiency by 30%.
• Improved search techniques by implementing distance and phonetic algorithms like Jaro-Winkler distance to match incorrectly typed search queries with the closest matching train name / station name.
• Analyzed user search and contribution parameters to provide insights for user retention strategies, contributing to a 5% increase in user retention.

Facial Recognition using Deep Learning

A research project to compare the performance of various multi class classifiers applied over Google's FaceNet based on a Convolutional Neural Network Architecture. Our end goal was to provide a system that makes it possible for a teacher to mark the attendance of a class by simply clicking a picture of the class.

A Tale of 3 Monkeys and 3 Humans

Theory of Computation design for river crossing puzzle

Fake Product Detection for E Commerce Websites

Machine Learning Pipeline for fake product detection for E-Commerce Websites

Key Value Database

Key Value Database with two variations showcasing - ArrayLists and HashMaps