arefin@skovde:~$ whoami
Shamsunnur Ibn Arefin
Data science student · former QA engineer
Five years in software quality assurance, now studying data science at the University of Skövde. I work on coursework and personal projects in machine learning, and I am building my first production web project, streetdudes.se. Aiming to built my next phase of the career as a Data Scientist and AI Quality Engineer.
/experience
System developer & kitchen assistant @ Streetdudes AB
May 2026 – present · Borås, Sweden · part-time
Working two roles at once: building the restaurant's systems as technical lead for web, and working shifts in the kitchen as an assistant cook.
Designed and built streetdudes.se, a bilingual (Swedish/English) website and online ordering system with Next.js, TypeScript, Supabase and Vercel. Online ordering is being prepared for launch.
Built the kitchen setup — a live order dashboard and receipt printing — plus an admin panel with sales analytics.
/education
Aug 2025 – Jun 2027 (expected)
MSc in Data Science @ University of Skövde
Skövde, Sweden · 120 credits
Master's thesis (30 credits) in spring 2027, topic open. Looking for applied machine learning on real device and sensor data.
Data Science Project (15 credits): remaining useful life estimation with knowledge distillation on NASA CMAPSS.
Graduated May 2019
BSc in Computer Science & Engineering @ BAIUST
Bangladesh
Industrial training at an ISO 9001:2015 certified software company.
/recent writing
all posts →/selected projects
all projects →streetdudes.se
Bilingual website and online ordering system for a restaurant in Borås, with a live kitchen order dashboard, receipt printing, and an admin panel with sales analytics.
github ↗Colon cancer tissue analysis
8-class histopathology image classifier comparing a CNN built from scratch with MobileNetV2 transfer learning — about 88% accuracy vs. 85%, with the main STROMA/COMPLEX confusion documented.
github ↗Student performance early warning
Predicts students at risk of failing from the UCI dataset (1,044 students). Earlier grades removed to avoid leakage; Random Forest reached 78.7% accuracy (F1 0.88).
github ↗/contact
Need help with something? I take on data science, testing, and web work, and I am open to consultation — from a quick second opinion on a dataset to building out a full project. Questions about something I wrote are welcome too. I usually reply within a few hours.