About Me
Cem Karışlı
Data Scientist & ML Engineer
Education
Master's Degree (Current)
Financial Data Analytics
University of Sussex
Master's Degree
Computer Science (Thesis)
Doğuş University
Bachelor's Degree
Industrial Engineering
Çukurova University
Projects
ML Classification Metrics Streamlit App
An application explaining Log Loss and Focal Loss metrics through formulas, charts, and interactive demos.
- Demonstrated the advantages, disadvantages, and usage scenarios of both metrics with interactive charts.
- Translated theory summarized from academic sources into clear, simple language.
- Published code and documentation on GitHub, deployed the app on Streamlit Cloud.
- Live demo: log-focal-loss.streamlit.app
Pneumonia Detection from Chest X-rays
Pneumonia diagnosis from X-rays using Xception-based CNN model with user interface.
- Designed PyQt GUI for X-ray upload and result display.
- Discussed model limitations by analyzing errors in class "0" due to data imbalance.
- Presented as a project for Computer Vision graduate course, selected as the best project.
Market Value Prediction for Investment Companies
Multiple linear regression model (R² ≈ 0.75) using financial variables.
- Built multiple regression model in SPSS using revenue, debt, and profit components.
- Trained on first 10 companies and tested prediction accuracy on 100 companies (R²≈0.75).
- Demonstrated possible ranking changes until 2030 through scenario analyses.
- Presented results as a report for my graduate course "Data Science and Business Intelligence".
TCDD Train Seat Availability Notifier
Selenium-based bot (Python) that sends automatic email notifications when seats become available.
- Periodically scraped train and seat data from TCDD website using Selenium.
- Sent HTML-formatted email notifications via SMTP when availability appeared.
- Configured script to run 24/7 on VPS with cron for uninterrupted operation.
- With this app, I easily found seats for friends wanting to travel from Ankara to Istanbul, even at the last minute :)
Visa Appointment Tracking Bot
Fully automated tracking bot with Oxylabs proxy, 2Captcha, and Telegram integration. (Python)
- Bypassed Cloudflare using undetected-chromedriver + Oxylabs proxy.
- Solved visual CAPTCHAs with 2Captcha API; implemented retry logic for errors.
- Sent results and logs to Telegram bot via cron job running every minute.
- Installed complete setup on my Raspberry Pi for continuous operation.
Articles
Detecting Credit Card Fraud with Anomaly Detection
Using anomaly detection algorithms for credit card fraud detection, data preprocessing techniques, and machine learning models for real-time fraud detection systems.