Available for opportunities

Sabbir RayhanMahee.

I'm a

I build production-grade web applications end-to-end — from database schema and REST API design to calm, responsive interfaces — and apply deep learning to real-world problems through research.

Khulna, Bangladesh

Full-Stack

Next.js · Node.js

500+

problems solved

Projects Shipped: 5+
Problems Solved: 500+
Technologies: 20+
Years Coding: 4+

01 About

Who I am & what I do

I'm a Computer Science student at CUET, Bangladesh, passionate about software engineering and building reliable, practical solutions to real-world problems.

I enjoy working across the software development lifecycle — from designing systems and architecting backend services to building polished, performant user interfaces. My primary stack includes Next.js, React, TypeScript, Node.js, PostgreSQL and Prisma.

I have experience developing and deploying full-stack applications with payment integrations, role-based authentication and database-driven systems for real users. I care about clean, maintainable code and systems that are scalable and easy to evolve.

Alongside engineering, I work on applied machine learning and deep learning research in Python. My B.Sc. thesis detects AI-generated Bangla speech by fine-tuning a self-supervised speech model, and my IEEE conference paper (under review) studies how scaffold-aware evaluation exposes generalization gaps in drug solubility prediction.

Beyond individual projects, I've collaborated on university software projects — from requirements and system design to implementation, testing and deployment — while continuously strengthening my engineering fundamentals.

Location

Khulna, Bangladesh

University

CUET, Bangladesh

Focus

Software Engineering · Full-Stack

s.rayhanmahee@gmail.com

Sabbir Rayhan Mahee

Software Engineer · CUET

Education

Mar 2022 – Aug 20263.28 / 4.00
Chittagong University of Engineering and Technology (CUET)
B.Sc. in Computer Science and Engineering
2018 – 2020GPA 5.00 / 5.00
Noapara Govt. College
Higher Secondary Certificate — Science

02 Skills

My technical toolkit

Technologies and tools I use to design, build and ship production apps — and to train and deploy machine learning models.

Frontend

Next.jsReactTypeScriptTailwind CSSShadcn UIZustandReact Hook FormBetter-AuthFirebaseZodFramer Motion

Backend

Node.jsExpress.jsREST APIJWT AuthenticationbcryptSSLCommerz

AI & Machine Learning

PythonMachine LearningDeep LearningPyTorchHugging Face Transformersscikit-learnXGBoostNumPyPandasSHAPStreamlitGradio

Database & ORM

PostgreSQLPrisma ORMMongoDBMySQLNeon

DevOps & Tools

GitGitHubVercelRailwayLinuxVS CodePostman

Competitive Programming

C++Data StructuresAlgorithmsProblem Solving

03 Experience

Work & experience

2023 – Present

Academic

Full-Stack Developer

University Group Projects — CUET

Collaborated in a team of 3–4 developers on multiple university-assigned full-stack projects. Took ownership of backend architecture, API design, and database schema while contributing to frontend development. Practiced agile workflows using GitHub for version control and code review.

Led backend development for CUETFinders — a real-world campus platform actively used by CUET students
Designed RESTful API architecture with proper authentication, authorization, and error handling
Collaborated on frontend-backend integration using Git branching strategies and pull request reviews
Delivered projects under academic deadlines while maintaining code quality and documentation

Node.jsMongoDBExpress.jsJavaScriptGitGitHub

2024 – Present

Self-Directed

Independent Developer

Personal & Freelance Projects

Independently designed and shipped production-grade full-stack applications using modern technologies. Focused on learning industry-standard practices including type safety with TypeScript, ORM-based database management, payment gateway integration, and cloud deployment.

Built and deployed Planora — a live event management platform with SSLCommerz payment integration
Implemented JWT-based authentication, role-based access control, and refresh token systems
Deployed full-stack applications on Vercel (frontend) and Railway (backend) with CI/CD via GitHub
Integrated Prisma ORM over PostgreSQL for type-safe database operations

Next.jsTypeScriptPrisma ORMPostgreSQLVercelRailway

04 Research

Research & applied AI

Machine learning and deep learning work, from problem framing and rigorous evaluation to deployed, publicly usable models.

B.Sc. ThesisCUET · 2026

Deepfake Detection from Bangla Audio using Deep Learning

Synthetic voice cloning is becoming increasingly realistic, while Bangla, spoken by over 230 million people, remains under-represented in audio anti-spoofing research. This thesis builds a deep learning system that distinguishes genuine Bangla speech from AI-generated speech, and deploys it as a public web application.

Fine-tuned Microsoft's self-supervised WavLM speech encoder, selectively unfreezing the upper transformer layers and adding a LayerNorm–GELU classification head.
Segment-level inference: recordings are resampled to 16 kHz, split into fixed windows, and per-segment probabilities are aggregated into a clip-level verdict with confidence.
Robust audio pipeline supporting WAV, MP3 and AAC/M4A uploads as well as live in-browser microphone recording (1–60 s).
Deployed on Hugging Face Spaces with a Gradio interface for real-time genuine-vs-fake detection.

PythonPyTorchHugging Face TransformersWavLMLibrosaGradio

Try the live demo Model & code

IEEE Conference PaperUnder review · 2026

Scaffold-Aware Drug Solubility Prediction

“Scaffold-Aware Drug Solubility Classification: Quantifying Generalization Gaps in Machine Learning Models with SMOTE and Extended Molecular Fingerprints on AqSolDB”

Aqueous solubility determines whether a drug can be absorbed by the body, so predicting it early saves costly lab synthesis. This work classifies molecules as High, Medium or Low solubility directly from SMILES strings, and shows how conventional random splits overstate real-world model performance.

Bemis–Murcko scaffold split keeps test molecules structurally novel, revealing that random splits overestimate accuracy by 3.91% and ROC-AUC by 5.62%.
525-feature representation combining 13 RDKit physicochemical descriptors with dual ECFP4 + ECFP6 molecular fingerprints.
SMOTE rebalancing for the minority high-solubility class (10.6% of data) with a tuned XGBoost classifier.
SHAP explainability identifies LogP (lipophilicity) as the dominant predictor, consistent with established chemistry.

PythonXGBoostscikit-learnRDKitSMOTESHAPStreamlit

Open the live app Source code

05 Projects

Things I've built & shipped

View all on GitHub

012026Featured

Planora

Event Management Platform

A full-stack event management platform where users can create, manage, and join events. Features role-based access (User/Admin), public & private events, SSLCommerz payment integration, invitation system with host approval/rejection, and smart payment retry flow.

Next.js 16TypeScriptPrisma ORMPostgreSQLSSLCommerzTailwind CSS+4

Live demo Frontend Backend

Details

022026Featured

Dragon News

Responsive News Portal

A fully responsive news portal with category-based browsing, live news ticker, and social authentication. Features Google/GitHub login via Firebase, a mobile slide-in drawer for quick access to social links and widgets, and a clean, newspaper-inspired UI built for readability across every screen size.

React 19ViteTailwind CSS v4DaisyUIFirebase AuthReact Router+1

Live demo Source

Details

032026Featured

SkillBridge

Tutoring Marketplace

A tutoring marketplace where students browse tutor profiles, book sessions, and manage bookings from a personal dashboard. Features separate role-based dashboards for Students, Tutors, and Admins.

Next.js 15TypeScriptTailwind CSSbetter-authAxiosREST API+1

Live demo Frontend Backend

Details

2025

CUETFinders

Campus Lost & Found Platform

A web platform for CUET students to report and locate lost items on campus. Features search, filtering, item status tracking, and a fully mobile-optimized interface.

Node.jsExpress.jsMongoDBJavaScript

2025

E-Commerce Website

Full-Stack Shopping Platform

A fully functional e-commerce site with product listings, shopping cart, and secure checkout flow. Built with Node.js, EJS templating, and MongoDB for dynamic product management.

Node.jsEJSMongoDBJavaScript

06 Competitive Programming

Sharpening the fundamentals

Active across multiple platforms since 2022. Problem-solving keeps me sharp on algorithms, data structures and optimization — directly improving how I design backend systems.

Total problems solved

500+

across all platforms

CF rating

Max streak

Active since

“Competitive programming isn't just about solving problems — it's about thinking systematically and writing code that works correctly under constraints.”

07 Contact

Let's build something great together

I'm currently open to new opportunities — full-time roles, internships, freelance projects, or just a good conversation. My inbox is always open.