Projects
Academic and software projects. Each card links to the code on GitHub.
A command-line tool that grades handwritten PDF submissions against a master solution with a two-stage multimodal LLM pipeline served through NVIDIA NIM. It saves checkpoints, retries on rate limits and writes a summary CSV for each batch.
A progressive web app for a small group of friends to log workouts against their own split days. It estimates calories from meal photos with an LLM and syncs step counts from iOS Shortcuts.
A pipeline that uses BERT and GPT-2 to classify network flows from the CSE-CIC-IDS2018 intrusion detection dataset, in binary and multi-class settings.
Logistic regression with bagging and stacking ensembles, PCA, an EM algorithm for Poisson mixture models and a feed-forward neural network with batch normalization, dropout and the Adam optimizer, all built from scratch in NumPy. The network is trained on Fashion-MNIST.
A full-stack document management app with file flow, sharing, digital signatures and verification. My work focused on the back end.
Implemented AES, Diffie-Hellman and RSA with socket communication. Demonstrated a buffer-overflow attack and teaching malware in Docker.
Built a multithreaded client-server program with sockets. Configured NAT and ACLs on VLANs. Simulated wired and wireless mobile networks and implemented TCP Adaptive Reno congestion control in NS3.
Practiced Bash scripting and thread synchronization with pthreads. Added system calls and a scheduler with lottery and round-robin queues to the xv6 operating system.
A full-stack anime database with listings, discovery and forum discussions. My work focused on the front end and full-stack integration.