Statistics · Dokuz Eylül · İzmir
Statistics student · Data Mining & Machine Learning
Open to internships
01 — Sample
I'm Gökhan, a final-year Statistics student at Dokuz Eylül University. When I look at data, I read the distribution before the story: separating signal from noise, and being able to explain why a model decides what it decides. I build end-to-end projects, from credit risk and plate recognition to explainable models and LLM agents. Now I'm looking for an internship to test that approach on real data.
02 — Models
Each cluster in the cloud is a project. Each one answers the same questions: what was the problem, how did I solve it, and how far can the result be trusted?
05 projects
A YOLOv8 I fine-tuned finds plates at 0.94 mAP50; OCR reads them and a local LLM sums them up.
Python · YOLOv8 · OpenCV · Tesseract OCR · Ollama · +1
A transparent model matched the black boxes: EBM won at AUC 0.809 and explains every decision.
Python · pandas · scikit-learn · XGBoost · LightGBM · +5
Built from scratch in 8 days: a banking assistant with 5 specialist agents, RAG and safety enforced in code.
Python · Ollama · qwen2.5:7b · nomic-embed-text · SQLite · +2
Calibrated XGBoost at AUC 0.893, after catching 5 data leaks that 20+ public notebooks missed.
Python · pandas · NumPy · scikit-learn · XGBoost · +6
Graduation project: statistically testing how ESG scores relate to firm performance.
Pre-registered · H₀ / H₁ written before results
Pulls store, date and total from receipt photos with EasyOCR and regex, stores them in SQLite without duplicates, and charts spending by store and month in a Flask app with Chart.js.
Python · EasyOCR · OpenCV · Flask · SQLite · Chart.js
View code03 — Variables
Each point is a skill. The x axis runs from statistics to engineering; the y axis counts the projects that used it.
ŷ = 1.23 + 0.84·x. Slope 0.84, r 0.23, R² 0.05, n 24.
y is the number of projects that used each skill, counted from the READMEs rather than estimated. The line is a live OLS fit of the visible points and refits as you filter. Hollow points are interests that haven't become a project yet.
04 — Time series
Started my BSc in Statistics in İzmir. The first observation in the sample.
Caught 5 data leaks in 148,670 records; calibrated XGBoost reached AUC 0.893.
Built a banking assistant with 5 specialist agents and RAG from scratch.
A transparent EBM proved statistically on par with black-box models: AUC 0.809.
Fine-tuned YOLOv8 to 0.94 mAP50, then chained OCR, regex and a local LLM end to end.
Started my graduation project on how ESG scores relate to firm performance. In progress.
Today
Expected to finish my BSc in Statistics. The biggest source of variance at this point: which team, and which data, I'll be working with.
Forecast
Everything after today is a forecast, not an observation. The interval is wide for now; the right internship narrows it.
05 — Hypothesis
H₀: We haven't met yet.
If your team has room for an intern who works with data, a short email is enough. Let's reject the null hypothesis together.
Time in İzmir