B.Sc. · Computer Science and Engineering (CSE)
Computer Science graduate (CGPA 3.65/4.00, Daffodil International University) with hands-on exposure to operating systems, networking, and database administration through academic and volunteer work. Designed and hardened a hybrid DNS infrastructure (BIND9 + Webmin) achieving ~18ms response time and 100% query accuracy under load, and independently configured Ubuntu environments and resolved technical issues on 55+ laptops as an on-site volunteer for the ICPC Asia Dhaka Regional Contest 2024. Holds formal training in MySQL, SQL Server, and Oracle database concepts and an AWS Academy Cloud Foundations certification. Comfortable with Windows and Linux environments, Microsoft Office, and clear technical documentation, with strong attention to detail and a quick learning curve on new tools and systems — looking to bring this foundation to an IT Support/IT Operations internship in the financial services sector.
Final year project generating Bangla MCQs from text/images using fine-tuned LLMs, with a Streamlit quiz UI.
My final year project: a system that auto-generates Bangla-language multiple-choice questions from text and images. I fine-tuned and compared multiple LLMs (Qwen3-VL-8B, BanglaT5, Gemma, Pixtral/Ministral 12B) using LoRA/PEFT on the Noirbittik Bangla dataset via Kaggle — BanglaT5 proved best for explainability, Qwen3-VL-8B for generation quality. I built a Streamlit UI with quiz interaction, Jaccard-similarity checks against training data to catch leakage, history/records tabs, screen recording, and CSV/JSON export, deployed via Cloudflare tunnel. I also fixed several critical training bugs (data leakage, global variable misuse, missing eval metrics, checkpoint selection) and separately debugged a Mistral-7B notebook through SFTTrainer/FastLanguageModel compatibility issues.
Note: the repo mcqcreator is currently empty on GitHub — you'll want to push your code before linking it, or use the repo where the actual work lives if it's elsewhere.
2022 – 2026
B.Sc. · Computer Science and Engineering (CSE) · 3.65 / 4.00
2019 – 2020
Higher Secondary Certificate (HSC) — Science · 5.00 / 5.00
2017 – 2018
Secondary School Certificate (SSC) — Science · 4.94 / 5.00
Lets users link bKash, Nagad, Rocket, and Upay accounts and transfer funds between them from one app.
Mobile financial services in Bangladesh operate as isolated silos — users can't move money between bKash, Nagad, Rocket, or Upay without juggling separate apps and logins. Uni-Pay solves this with a single Flutter app that lets users view all linked account balances in one dashboard, transfer funds directly between any two supported platforms with PIN confirmation, track transaction history centrally, and check live top-20 cryptocurrency prices via the CoinGecko API. I built the onboarding flow, dashboard, cross-platform transfer screens, transaction history, and crypto price screen, structuring the app around a clean screens/ + services/ architecture. It's a UI/UX and mobile-development exercise (built for the Mobile Application Design Lab course) — it doesn't move real funds, but demonstrates a complete cross-platform-transfer UX end to end.
A real-time 2D boat survival game rendered from raw OpenGL primitives — no engine, no sprites.
A Computer Graphics Lab project where I built a 2D survival game entirely with OpenGL fixed-function primitives (GL_QUADS, GL_POLYGON, GL_TRIANGLE_FAN) instead of a game engine or sprite assets. The player steers a boat through a scrolling river, dodging rocks and floating logs, with AABB collision detection, a health/score system, difficulty scaling, and a 3-charge speed boost. I implemented sinusoidal wave animation for the river surface, modular arithmetic for seamless parallax tree scrolling on both banks, real-time animation via a GLUT timer, and full screen-state management (start screen, gameplay, game over, restart). The whole thing runs on Python with PyOpenGL and FreeGLUT.
Classifies vegetable photos into 15 categories with 99.97% validation accuracy, deployed via Streamlit.
A computer vision project that fine-tunes a ResNet18 model to classify vegetable images into 15 classes, trained on roughly 18,000 images and reaching 99.97% validation accuracy. I built a Streamlit front end where users upload a photo and get an instant prediction, with the trained model auto-downloaded from Google Drive at runtime so the repo stays lightweight. This has been a centerpiece project when applying for AI/ML Engineer (Computer Vision) roles.
A desktop banking app for account management, transactions, and balance inquiries, built with Java Swing and MySQL.
A desktop-based Java application that streamlines core banking operations: creating and updating accounts, performing deposits/withdrawals/transfers, checking balances in real time, and viewing user-specific account details. The GUI is built with Java Swing for an interactive desktop experience, backed by a MySQL database via JDBC for persistent, relational storage of account and transaction data.
2026
Cloud Foundations
2025
MySQL / Oracle / SQL Server
Bangladesh Computer Council (EDGE Project)
2024
ICPC Asia Dhaka Regional Contest
2026
YKSG-1 Freshers' Orientation