AI Research & Tools
My recent work explores the intersection of Generative AI, Large Language Models (LLMs), and systems engineering. I build solutions that bridge the gap between natural language and complex technical systems.
SlideDeck AI
A generative AI tool designed for the co-creation of presentation decks, streamlining the transition from conceptual ideas to structured slides using LLMs. Won 3rd place in the Llama 2 Hackathon with Clarifai.
KodeAgent
The minimalistic approach to building AI agents. KodeAgent is a small (~2,000 lines of pure Python) engine designed to be the robust reasoning core inside your larger system, not the entire platform.
Drive Sonar
Fast, lightweight disk-space analyzer that runs without admin privileges! Optimized for Windows; runs on Linux and macOS as well. Built with Rust/Tauri/React.
Open Data
I believe in open research and have authored several datasets to enhance AI capabilities in niche domains, specifically in code and Bengali literature.
CodePrompt Dataset
A resource developed to enhance and evaluate the capabilities of AI in code-related tasks and prompt engineering. It consists of hand-crafted coding problems for evaluating LLMs.
Paper Dataset Get it →
GanitPrabaha
A specialized dataset for Bengali math reasoning, designed to improve the mathematical capabilities of LLMs in the Bengali language.
Dataset Get it →
Bangla Nirdeshabali
A question-answer dataset in Bengali covering literature, culture, and geography, created to support instruction-following in Bengali LLMs.
Dataset Get it →
NLU Dataset for IBNs
A synthetically generated dataset to train Natural Language Understanding models for Intent-based Networks (IBNs).
Paper Dataset Get it →
Bengali Poems Dataset
A curated subset of the Bangla Sahitya dataset specifically containing Bengali poems, made available on Kaggle for communal research and use.
Dataset Get it →
Bangla Sahitya
A non-exhaustive collection of Bengali literary works by more than 35 authors, spanning 12 centuries.
Dataset Get it →
Software Engineering & Interactive Labs
Before focusing on AI research, I developed a suite of interactive, web-based laboratories and tools designed for global engineering education. These projects focused on providing hands-on practice in complex technical subjects.
Software Engineering & Network Virtual Labs
A suite of interactive, web-based laboratories designed to provide hands-on practice in ER modeling, UML diagrams, and network simulations (NS-2/NS-3). These labs have been used by students globally to bridge the gap between theory and practice.
Education System Design Software Engg Lab → Network Lab →
Ns2web
A standalone web-based interface for performing remote simulations and analysis with NS-2. It features a guided scenario generator and exports research-quality plots (EPS, PDF, PNG).
Simulation Try it →
The ONE Knowledge Base
A searchable archive of the Opportunistic Networking Environment (ONE) simulator community mailing lists. Built using Django and Haystack to provide an easy-to-use research resource.
Research Tool GitHub →
Software Engineering Lite
An unofficial port of the Software Engineering Virtual Lab for Android devices, designed to make educational content accessible on the go.
Android GitHub →