My Framer Site

Software Projects Overview

My software projects span over both my university and work experiences. I am excited to share my favorites with you.

BoMination: Automatic BoM to cost sheet application

BoMination is an end-to-end intelligent system designed to automate the extraction and processing of Bills of Materials from engineering schematics. Developed for use in a control panel manufacturing environment, this tool significantly reduces manual labor and error-prone processes traditionally involved in reading and compiling BoMs. At its core, the system uses a YOLO-based object detection model trained on real schematics to accurately locate BoM tables embedded within PDF documents.

Once the BoM region is identified, the tool applies advanced computer vision techniques and OCR preprocessing to segment the table and extract structured text data. Super-resolution models enhance image clarity for improved text recognition, while Tesseract OCR is used in a two-pass system to refine accuracy, especially on complex or low-quality schematics. The extracted data is then exported into a formatted Excel spreadsheet, where users can optionally review or make adjustments before continuing to the next stage of the automation.

After manual verification, the pipeline resumes by finding manufacturer part numbers and accessing OEMsecret's BoM tool using selenium to identify the best available price and lead time. This semi-automated procurement process delivers a final Excel output optimized for purchasing decisions, cutting what used to be an hours-long process down to just minutes. The tool is modular, fast, and designed to be integrated into real engineering workflows, offering meaningful gains in both efficiency and reliability.

Github Repo

Analysis of Zero Trust Architecture Frameworks

To evaluate the practical deployment of modern cybersecurity paradigms, I conducted an experimental and theoretical analysis of three leading Zero Trust Architecture (ZTA) models: Agent Gateway, Agentless Gateway, and On-Device Agent. This project, completed as a final paper for ECE 547: Security Engineering, benchmarked each model's performance in terms of latency, scalability, resource efficiency, and fault tolerance.

Using a custom Python-based simulation environment with Redis integration for session management and rate limiting, I tested each model under real-world traffic conditions. The results demonstrated the trade-offs of centralized vs. decentralized architectures and helped identify the ideal use cases for each system from high-traffic corporate environments to resource-constrained IoT networks.

This work not only aligned with current theoretical literature but expanded upon it by identifying key deviations and proposing future integration with blockchain-based trust management. The analysis contributes a practical lens to ZTA evaluation and provides a framework for secure, scalable cybersecurity design in dynamic environments.

Github Repo

Resume Evaluation System with NLP & Scoring Metrics

To explore the real-world application of natural language processing in recruitment, my team and I developed a tool that evaluates and optimizes resumes against specific job descriptions using both keyword matching and large language model (LLM) prompt engineering.

Our system processes resumes and job descriptions to compute a relevance score based on a manually curated list of 76 industry buzzwords. We then used multiple prompt variations with GPT-4o-mini to tailor resumes for different roles (internships, entry-level, and senior SWE jobs). We evaluated these revised resumes using a custom scoring formula and human reviewers from the software industry.

In our human evaluation phase, professionals noted that while AI-generated resumes occasionally hallucinated content, they consistently aligned more closely with the job requirements and would be strong candidates with light edits. The project demonstrates how LLMs can accelerate resume optimization at scale, while highlighting the importance of human oversight. Built with spaCy, scikit-learn, and matplotlib in a team of three for a final project in COMPSCI 383.

Github Repo