Pysmo
May 2023 to Present
A modular Python framework for seismology
Projects
Reinforcement Learning for High-Frequency Trading
September 2020 to March 2021
Designed an environment for training RL agents on high-frequency trading data. Developed a DDQN agent to leverage this HFT data to take intelligent decisions.
LinkedIn Network Analytics
September 2020 to December 2020
Analyzed how our LinkedIn network has changed in the post-COVID era as compared to the pre-COVID era.
Fine-Tuning BERT for Text Summarization
May 2020
Fine tuned BERT for extractive text summarization
Analyzing spread of COVID-19 using Graph Neural Networks
March 2020
Developed an end-to-end pipeline for analyzing the spread of COVID-19 using Graph Neural Networks. Users can process COVID-19 data, form a graph out of it and apply various Graph Neural Network algorithms on it.
Domain Adaptation using CycleGAN
January 2020 to March 2020
Developed a CycleGAN architecture for generating real-world images from simulated images to reduce the domain gap between real-world data and simulated data.
Citizens Police Data Project
September 2019 to December 2019
In this project, we analyzed the trends after CPDB (Citizens Police Database) went public and compared it with the trends before the release of CPDB.
Simulation Environment for Development and Testing of Autonomous Learning Agents
August 2018 to May 2019
Developed a simulation environment for autonomous driving agents which has customizable training and testing parameters, customizable car parameters, simulated pedestrian and vehicular traffic and an easy to use road editor. The environment also provides a plug and play interface for supervised and reinforcement learning agents.
Password Cracking and Strengthening Tool
October 2018
Developed a password cracking and strengthening tool which allows users to crack password using Brute Force, Dictionary Attacks and Collision Attacks. The tool also allows the user to adjust different parameters to suit their needs. The tool provides strengthening features as follows: • Users can generate strong random passwords • Users can generate strong easy to remember passwords • Users can generate a stronger version of their existing passwords(Improvements to existing password)
k-Shot Facial Recognition System
January 2018 to April 2018
Developed a k-Shot Facial Recognition System which uses a combination of Dlib, Residual Neural Network and a Fully Connected network to learn from a small number of samples. The system produces a 100% accuracy with training samples per subject as small as 3, as long as the output classes are limited to a small number.
Context Aware Searching
January 2018 to March 2018
Created a system which predicts related keywords based on the input query by using Hidden Markov Models and a Neural Embedding Network. The system can be used to better understand user queries and display information based on the context of the query instead of simply presenting information based on the query itself.
Credit Card Fraud Detector
June 2017
Developed a credit card fraud detector which is trained on a skewed dataset of transactions. It uses anomaly detection to separate out fraudulent transactions from normal transactions.
Poetry Writer AI
December 2016 to January 2017
Developed a Poetry Writer AI which learns to write poetry by using Hidden Markov Models.
Movie Recommender System
December 2016
Developed a Movie Recommender System which forms the database by scraping information from the web and then learns user preference to recommend new movies.
AI based Tic Tac Toe
September 2016 to October 2016
Developed a Tic Tac Toe game with a graphical user interface where the user can play against an AI agent which uses Minimax algorithm to perform the best action at every step.
Mall Pro
December 2015 to January 2016
Developed a server based android application which can be used to get information and listing of malls. The application provides real-time listing of offers and discounts offered by store outlets in malls.