Richik SCExplored low-computation approaches in deep policy gradient algorithms such as gradient-based entropy scheduling and gradient signal processing to accelerate learning and maintain long-term training stability.
reinforcement learningpythonpytorch
Hi, I'm Richik.
I'm a software engineer and graduate student at Northeastern University. I graduated with my bachelor's degree in Computer Science from the University of Wisconsin-Madison in 2024.
Formerly, I was a Technology Development Program Intern at Optum, working with a site reliability engineering team.
Prior to that, I interned at PTC on the Vuforia AR team.
I was also a teaching assistant for the undergrad Operating Systems course at both UW-Madison and Northeastern.
View my projects
Projects
CNN Image Classifier From Scratch
A CIFAR-10 image classification model implemented using only elementary matrix/tensor operations. Implemented modular 2D convolution, max-pooling and spatial batch normalization layers with composable backpropagation. Achieved 67.1% classification accuracy on test dataset after only 10 seconds of training.
computer visionmachine learningpythonpytorchSpotify Streams Predictor
Developed and trained a neural-network based model to predict the number of Spotify streams of a song based on track attributes. Used Pandas to load, clean, and preprocess training data, using Tensorflow and Keras to develop the model. Evaluated various loss functions and tuned hyperparameters to improve model performance.
machine learningdata sciencepythontensorflowpandasA low-level implementation of an RPC client library and server in C using TCP sockets. Multi-threaded implementation handles up to 100 concurrent clients accessing a shared datastore.
distributed systemsclinuxhttps://mercantile.richiksc.me Mercantile is a product inventory management system built using Spring Boot and Java for the backend, and a VueJS frontend. Deployed to Google Compute Engine.
vuejsspring bootjavajavascriptwebappcloud