Different languages. Shared context.
A multi-task pipeline that understands, classifies and prioritizes support tickets in three languages.
Explore projectAI / ML / Data Science
From data to insight. From ideas to systems.
I work on machine learning, natural language processing and reliable backend systems. My focus is turning models into products that solve real problems.
Konya, TürkiyeKTO Karatay
Natural Language Processing✳Machine Learning✳Backend Engineering✳Automation
Selected work
Each project answers a different question.
StockFlow
Multi-warehouse inventory and asset management, built around data integrity and authorization.
Explore projectThink better with your PDFs.
An AI-powered, all-in-one PDF assistant, aiming to make working with documents more meaningful.
Explore projectMore experiments
About
Start with curiosity. Learn by building.
Mustafa Arıkan↗
I am a final-year Computer Engineering student at KTO Karatay University. Across three internships at Deva Yazılım, I built multilingual ticket classification, flood early-warning and full-stack applications. I enjoy the full journey from understanding data to evaluating and deploying models.
⌕Understand the data first
≈Evaluate with the right metric
↗Ship a working system
AI, machine learning, data science and automation
Experience
Beyond the theory.
Three internships. Practical experience from data to deployment.
Download CVJune – August 2026
Computer Engineering Intern
Deva Yazılım Entegre Çözümler
End-to-end NLP pipeline for English, German and Turkish support tickets.
- Multi-task mBERT over 32.7K tickets, predicting type, queue, category and priority.
- MLflow experiment tracking; mitigated language confounding with stratified, customer-grouped splits.
- FastAPI and Kubernetes deployment with Locust, Prometheus, Grafana and drift monitoring.
September – December 2025
Computer Engineering Intern · Part-time
Deva Yazılım Entegre Çözümler
Flood early warning, from simulation to machine learning.
- EPA SWMM / PySWMM simulations from DEM data and three-level flood risk classification.
- 30–45 minutes of early warning at an 81% success rate.
July – September 2025
Computer Engineering Intern
Deva Yazılım Entegre Çözümler
InternTracker: an internship platform for tasks, goals and journals.
- 7 controllers and 55 Razor views with ASP.NET Core MVC, EF Core and SQL Server.
- Admin, mentor and intern roles; access-controlled file sharing and dashboards.
My toolkit
The problem comes before the tool.
✳
AI and data science
NLPComputer VisionTime-Series ForecastingFeature EngineeringModel Explainability
⌘
Models and tools
PyTorchTensorFlowHugging Facescikit-learnXGBoostMLflowpandasNumPy
⌁
Application engineering
PythonC#TypeScriptFastAPIASP.NET CoreReactSQL ServerSQLite
↗
Operations and delivery
DockerKubernetesGitHub ActionsPrometheusGrafanaLokiGit
Out in the open
Learning, building, sharing.
github.com/MustafArikan
10Public repositories12Stars received
The contribution calendar is not available in this view yet. Visit my GitHub profile for the latest work.
Support-Ticket-Classification-and-Priority-Prediction
Build a system that automatically classifies incoming support requests (email, form, chat) into categories (Technical Issue, Billing, Refund, Account, General Info), predicts priority (Low/Medium/High/Critical), and explains decisions using SHAP/LIME, deployed as an observable service with CI/CD on Docker + Kubernetes.
Jupyter Notebook1
Plant-Disease-Detection-with-Deep-Learning
Developed a deep learning–based image classification system to automatically detect tomato plant diseases from leaf images. The project utilizes transfer learning with MobileNetV2 to achieve high accuracy while reducing computational cost.
Jupyter Notebook1
video-game-hit-predictor
An Ensemble Machine Learning pipeline (Keras DNN + XGBoost) predicting video game hits using sales and review │ data, featuring SMOTE, target encoding, and automated hyperparameter tuning.
Python1
star-type-classification
An end-to-end Machine Learning pipeline using KNN, Decision Trees, and Random Forest to classify stars into 6 distinct types based on physical characteristics.
Jupyter Notebook1
Repository primary languagesJupyter Notebook 6Python 1HTML 1JavaScript 1
Updated: 13 Nov 2026
Education
KTO Karatay Üniversitesi
B.Sc. Computer Engineering · Final year
Expected graduation: 2027
Machine Learning, Deep Learning, Data Structures and Algorithms, Probability and Statistics, Signals and Systems
Contact
What can we build together?
Leave a message to talk about AI, data or software.
arikanmustafa@gmail.com