Center for Entrepreneurship, University of Southern Mississippi · Hattiesburg, MS

Build and maintain the Center's web platforms and internal tools, and develop and manage AI-agent workflows that support its entrepreneurship programs.

Ma Research Lab, University of Southern Mississippi · Hattiesburg, MS

Building ML pipelines to automatically characterize and speed up Atomic Force Microscopy imaging of block copolymers.

• Trained a custom U-Net to segment nanoscale features in AFM images, a 35% improvement over conventional image analysis
• Benchmarked CNN, transformer, and GAN super-resolution models to cut AFM scan time by up to 3.7x (manuscript in preparation)

BIONET Project (Euro-African Biomonitoring Network)

Built and deployed a TYPO3 website for a network assessing environmental exposure in populations.

• Developed the CMS-driven site and deployed it on Azure with Apache and MySQL

Cyber Innovations Lab, University of Southern Mississippi · Hattiesburg, MS

Built ML and deep-learning pipelines for AES-128 key recovery via electromagnetic side-channel analysis on the ASCAD dataset.

• Trained and benchmarked CNN, ResNet, SVM, and feature-selected Random Forest classifiers against EM traces, reaching Rank-0 key recovery
• Work published at the 34th SEDE Conference (AI Track), 2025 (Springer CCIS)

Quantum Undergraduate Network

Co-founded a global community that helps undergraduates get into quantum information science, and solo-built its web platform.

• Designed, built, and deployed the full React and Next.js platform on Vercel for talks, resources, and networking

Nepal Health Frontiers

Lead the technical infrastructure and digital presence of a healthcare NGO working on non-communicable diseases, mental health, and healthy-lungs campaigns.

• Maintain the website and email systems and publish the monthly research roundup

A platform that automates the federal benefits lifecycle, screening, enrollment, and renewal, so people don't lose SNAP, Medicaid, and Medicare to processing errors and missed deadlines, and states cut the error rates that trigger federal penalties. A startup I co-founded; winner of USM's Golden Idea Pitch and Checkpoint programs.

An efficient natural-language-to-shell model that runs entirely on CPU. I fine-tuned and 4-bit quantized Qwen2.5-Coder-1.5B (LoRA, GGUF Q4_K_M) into a 941 MB model with roughly 1-second latency. It scores 0.620 on the 300-task InterCode-ALFA execution benchmark, a statistically significant lift over the untuned baseline and on par with an untuned 7B model.

A published JavaScript library that turns any image into an interactive, physics-driven particle cloud. Zero dependencies and plug-and-play. It powers the photo on this site's home page.

Local-first semantic search for your shell history. It finds commands by what you were trying to do rather than the exact text you typed, using ONNX embeddings alongside FTS5 keyword search, an interactive TUI, and fully offline inference. Published on PyPI.

Machine-learning and deep-learning models (Random Forest, SVM, CNN, ResNet) that recover AES-128 keys from electromagnetic side-channel traces on the ASCAD dataset, reaching Rank-0 recovery. Published at the 34th SEDE Conference (AI Track), 2025.

A machine-learning pipeline that estimates suspended sediment concentration from Sentinel-2 imagery to improve flood-risk models for the Koshi River Basin. Funded by an Eagle SPUR grant and presented at MidSouth ASPRS 2026.

M. Poudel and N. Rahimi, “Machine Learning-Based AES Key Recovery via Side-Channel Analysis on the ASCAD Dataset,” 34th Software and Data Engineering (SEDE) Conference, AI Track, 2025. Springer CCIS, vol. 2720, pp. 334-352.

Super-resolution and segmentation for accelerated Atomic Force Microscopy of block copolymers (Ma Lab, University of Southern Mississippi).