Saroj Baral

About Me

I am Saroj Baral, an AI researcher and software engineer from Pokhara, Nepal, currently pursuing my M.S. in Computer Science at Middle Tennessee State University. I am passionate about exploring how computation helps us better understand and shape the world, with particular interests in machine learning security, AI fairness, and intelligent systems.

  • Education: Completed B.E. in Computer Engineering from Pulchowk Campus, Nepal with 78.8% (A grade). Currently maintaining a 4.0 GPA in my Master's program.
  • Research Experience: Working as a Research Assistant at AIbigdata Lab under Dr. Krishnat Poudel, contributing to NIH/NSF-funded projects on Medical LLM security and AI fairness.
  • Industry Experience: Worked as a Software Engineer at Growthzilla Inc for 10 months, developing survey analysis platforms and ML-based customer segmentation systems using React, Express.js, and D3.js.
  • Achievements: International Physics Olympiad participant (2019), National Physics Champion, and published researcher in IEEE conferences.
  • Personal Interests: In my free time, I enjoy solving mathematical problems, reading poetry, hiking, and exploring the intersection of technology and human behavior.

Projects

Here are some of the key projects I've worked on, ranging from AI research to full-stack development and computer graphics.

Lossless Text Compression Using GPT-2

Implemented a novel compression algorithm using GPT-2 language model achieving 15% better compression ratio than traditional methods like gzip. Applied techniques from Stanford research on neural text compression with custom probability modeling, reducing file sizes by 30% while maintaining 100% data integrity.

Python PyTorch NLP Transformers

Intelligent 29-Card Game AI Bot

Developed an advanced AI bot using Monte Carlo Tree Search (MCTS) variations for the traditional card game 29. Implemented sophisticated game strategy algorithms achieving 80% win rate against human players. Competition entry for Bhoos Entertainment, Inc. - placed in top 10 among 200+ submissions.

Python MCTS Game Theory AI

Academic Recommendation Letter System

Designed comprehensive web application for managing recommendation letters between faculty and students. Implemented secure user authentication, role-based access control, and automated email notifications. System now serves 500+ users with 99% uptime.

Django Python PostgreSQL Web Development

3D House Rendering Engine

Built custom 3D rendering engine from scratch using fundamental graphics programming techniques. Implemented OBJ file parser and low-level pixel rendering routines for complex 3D house models. Applied advanced lighting models, texture mapping, and camera controls for realistic visualization.

C/C++ OpenGL SFML Computer Graphics

Survey Analysis Platform

Built comprehensive survey analysis tool with interactive D3.js visualizations at Growthzilla Inc. Processed 10,000+ survey responses and reduced analysis time by 60%. Collaborated with design team to create intuitive user interfaces for data visualization.

JavaScript D3.js Data Visualization React

Publications / Papers

My published research work in machine learning and biomedical signal processing.

Personalized Stress Detection using a Lightweight Machine Learning Framework With Convenient Wrist-Worn Sensors (2024)

Authors: M. Hasan, S. Mithun, M. Mahmud, Saroj Baral, K. Poudel

Conference: 2024 IEEE Signal Processing in Medicine and Biology Symposium (SPMB), pp. 1-7

Abstract: This paper presents a lightweight machine learning framework for personalized stress detection using convenient wrist-worn sensors, contributing to the advancement of wearable health monitoring technologies.

Contact

Feel free to reach out for research collaborations, project discussions, or just to connect!