Loading Universe...
My Work

Projects & Innovations

A showcase of my technical journey through AI/ML research, full-stack development, and award-winning innovations. Each project represents a unique challenge and learning experience.

Edith GPT

Edith GPT

Web Dev

Next Generation AI Calendar: Edith is an intelligent calendar that not only plans your day but also considers your health, stress levels, and personal preferences. Through integration with multiple APIs, it suggests Spotify playlists for stress, plans academic tasks, offers horoscope-based guidance, allows conversational planning, and even suggests activities based on weather.

HTML5 CSS3 JavaScript Firebase ExpressJS NodeJS React TypeScript Git APIs
Nimbus AI(v1.0): Dark Matter

Nimbus AI(v1.0): Dark Matter

AI/ML

An advanced fine-tuned AI model designed for intelligent conversations, mathematical reasoning, fact-checking, and knowledge retrieval. NimbusAI integrates LLM fine-tuning, deep learning, and NLP to provide accurate responses, improve user interaction, and enhance automation.

Python TensorFlow PyTorch NLP Machine Learning Deep Learning LLM Fine-tuning Transformers

StorePro (Real Client Full Stack Website)

Web Dev

A comprehensive full-stack web application developed for a real client from Black Horse Pike, NJ. The system is divided into client and seller interfaces, managing appointments, schedules, services, inventory management, and customer relationships.

Flask Python HTML5 CSS3 JavaScript SQLite Bootstrap Jinja2

BattleX: PokemonTCG

Desktop

A comprehensive Pokémon Trading Card Game implementation built in Java using Swing for UI. Players battle using Pokémon, Trainer, and Energy cards, making strategic moves to defeat opponents. Features turn-based gameplay, card management, and strategic battle mechanics.

Java Swing OOP Game Development GUI Programming

Dementia Terminology Systems

Web Dev

A comprehensive medical terminology dictionary designed to simplify complex medical terms related to dementia. Provides clear definitions, medical explanations, and helps healthcare professionals and families understand dementia-related terminology.

Python Flask NLP Medical Computing SNOMED CT Healthcare IT

Bank Portfolio Manager

Desktop

A sophisticated Java-based desktop application with Swing interface for comprehensive bank account management. Features include transaction processing, loan management, account analytics, and secure data handling with encryption.

Java Swing SQLite OOP Financial Systems Data Security

iUtsa (Personal Portfolio Website)

Desktop

A modern, responsive portfolio website showcasing projects, skills, and achievements. Features dark/light mode toggle, interactive animations, contact forms, and comprehensive project documentation.

HTML5 CSS3 JavaScript Bootstrap Responsive Design Web Animation

FIFA Player Stats App

Desktop

A comprehensive desktop application for football enthusiasts to access detailed player statistics, recent performances, biographical details, and team information with an intuitive user-friendly interface.

Java Swing API Integration Data Visualization Sports Analytics

Graphing Calculator

Desktop

An advanced Java-based graphing calculator that allows users to input complex mathematical equations and visualize their graphs in real-time with an intuitive interface. Supports multiple function types and mathematical operations.

Java Swing Mathematical Computing Graph Visualization Algorithm Implementation

Hashimukh (Mental Health Assistant)

Desktop

A comprehensive mental health assistant designed to support individuals through meaningful conversations and facilitate access to professional care. Powered by GPT API, it engages users in empathetic conversations and provides secure appointment scheduling.

HTML5 CSS3 JavaScript NodeJS MongoDB GPT API Mental Health Tech
Nova-SLM-204 Parameters

Nova-SLM-204 Parameters

Research

NOVA-SLM is a 206M parameter language model specifically designed for scientific document understanding. Unlike general-purpose LLMs, NOVA-SLM introduces three novel architectural components that enable structure-aware processing of research papers. 🌟 Novel Contributions 1. SSAA - Semantic-Structural Adaptive Attention Unlike fixed sparse patterns (Longformer, BigBird), SSAA learns which tokens to attend to based on: Document section types (Abstract, Methods, Results, etc.) Structural importance scoring Cross-section compatibility 2. RSAPE - Research-Structure-Aware Positional Encoding Extends RoPE with document structure awareness: 50% RoPE: Standard rotary embeddings for sequence position 12.5% Section Type: Learnable embeddings for 11 section types 12.5% Hierarchy: Multi-level encoding (section → paragraph → sentence → token) 12.5% Citation Proximity: Distance to nearest citation (6 buckets) 12.5% Boundary Flags: Structural boundary markers 3. RTM - Research Trajectory Modeling Novel prediction heads for research supervision: Stage Classification (6 classes): Problem ID → Literature → Methods → Experiments → Results → Conclusion Risk Assessment (4 levels): Low → High research risk Quality Prediction (5 levels): Needs Improvement → Excellent

Python ML PyTorch Scikit-Learn SSAA RSAPE

GitHub Activity

Follow my coding journey and contributions on GitHub

Arnab Das Utsa

Arnab Das Utsa

@iUtsa

Follow on GitHub
20+
Repositories
500+
Commits
8
Languages
3+
Years Active

Contribution Graph

Consistent contributions throughout the year

Technologies I Work With

A comprehensive overview of my technical toolkit and expertise

Frontend

React
HTML
CSS
JavaScript
Bootstrap

Backend

Python
Flask
Django
Java
ExpressJS

AI/ML

TensorFlow
PyTorch
Scikit-learn
OpenCV

Database

SQL
MongoDB
PostgreSQL
SQLite

Tools

Git
Docker
AWS
Linux
VS Code

Interested in collaborating?

I'm always open to discussing new projects, creative ideas, and opportunities to create something amazing together.

Nimbus AI

Online

Hello! I'm Nimbus AI

I'm here to help you learn about Arnab's projects, skills, and achievements. Ask me anything!