Jul 2019 – Jul 2022 · 3 yrs · Bengaluru Area, India

May 2018 – Jul 2018 · 2 mos · Mumbai Area, India

• Project: Building a real-time logo transfiguration tool for compliance industry using deep learning techniques
• Reviewed several models: CycleGAN, CA-GAN, IcGAN, DeepFakes, DiscoGAN to get the best fit for the use case
• Implemented a Python script to parse the XML files containing the coordinates of logos in images to get training dataset
• Analyzed the training dynamics of IcGAN and CycleGAN on Google Cloud Platform under several synthetic settings
• Developed a robust model achieving Nash Equilibrium employing Instance Normalization and Patch Discriminator

Apr 2018 – Jul 2019 · 1 yr 3 mos · Remote

• Developed Trading Algorithms to deploy systematic investment strategies focused on US equity market inefficiencies
• Used advanced mathematical and statistical techniques to model and predict market movement for efficiently tr a ding equities
• Built robust model by seeking alphas from financial literature to achieve high Sharpe ratio and significant abnormal returns

Dec 2017 – Feb 2018 · 2 mos

• Achieved Gold level securing a rank of 13th out of 4000+ globally and was subsequently offered internship opportunity
• Developed algorithmic long-short market neutral trading signals on Websim Platform in Python and Expression language
• Submitted 406 Price-Volume, Fundamental Analysis and Analyst Estimate based strategies within 60 days

May 2017 – Jul 2017 · 2 mos · Gurgaon, India

• Project 1: Converting distorted feed from Fisheye lens camera to user-friendly view and surveillance
• Used Optimized Texture Mapping and Affine Transformation to dewarp 360 camera images in C++ using OpenCV
• Added hardware acceleration to the code by utilising Transparent API reducing CPU usage impressively by ~2x
• Project 2: Developing an end-to-end smoke detection module on a real-time video input
• Implemented Adaptive Gaussian Mixture Model and Statistical Ranking for background and foreground classification
• Detected the smoke pixel-wise applying Stationary Wavelet Transform and Colour Model on the foreground
• Tracked the smoke using Optical Flow Technique, Disorder and Connected Component Analysis with the accuracy of 83.4%