- Haier Europe is the 3rd largest white appliance company in the world. I worked for the Italy HQ office
- Engineered and productionised deep‑learning demand‑forecast models (AutoGluon, AWS SageMaker) to predict the next 12 months of sales for 26 countries, delivering a 30 % accuracy gain versus the prior baseline
- Built an automated workflow that extracts unstructured data from PDF invoices via LLMs (DeepSeek‑R1) orchestrated with LangChain, eliminating manual data entry and accelerating downstream reporting.
- Creating a full‑stack stock‑forecasting web app that pulls monthly price series from yfinance, enriches them with statistical technical indicators, and serves 3‑month forecasts from AutoGluon deep‑learning models via Python.
- The JavaScript front‑end renders interactive prices by retrieving data via the FAST API and indicator graphs, while an embedded LangChain‑powered chatbot delivers conversational equity insights.
- backend: Python, technical indicators, stocks data, JSON, PostgreSQL
- frontend: html, css, javascript, fastapi
- AI: time series deep learning, autogluon, LLM, langchain, chatbot, extraction, semantic search, AI agent
- Roadmap includes semantic search on news and agent‑driven RAG analytics for deeper market intelligence.
- Beko (formerly Arcelik) is the 20th biggest white appliance company in the world.
- Engineered and productionised deep‑learning demand forecast (AutoGluon, AWS SageMaker) to predict the next 12 months of sales for 50 countries, with a 40 % W-MAPE at the line level and improving planning accuracy.
- Designed and deployed a vessel‑scheduling model (AutoGluon regression, classification) that predicts monthly ETD/ETA with 80 % accuracy, accelerating container turnaround and improving on‑time delivery performance.
- Developed a statistical optimisation engine that analyses historic demurrage & detention data, establishes 95 % confidence windows for each route across 50 countries, and projects annual savings: US$250k.
- Built an end‑to‑end receipt‑processing pipeline that combines Google OCR/Speech APIs with Python (pandas) to convert images and audio into clean, structured transaction data for real‑time analytics.
- Drove data‑driven marketing strategy by implementing market‑basket analysis, customer segmentation experiments; Apriori association‑rule mining delivered > 80 % predictive precision for recommendations.
The University of Bonn is one of the largest research-based universities in the world. The THE ranking is 89 and the QS ranking is 201 in the world in 2023.
Being the best university in Germany in the field of mathematics, Uni of Bonn has also adjusted its computer science department to have a solid mathematical foundation. For this reason, many courses are built on combining heavy mathematical theory and practical coding skills, and the language of instruction is 100% English.
I took a total of 120 credits with a total of 14 courses and thesis+seminar from three different fields. All lecture details can be seen in files here with a summary of the content of each one of them. The programming languages in classes are mostly Python and C++. The transcript and diploma supplement can be seen here
intelligence systems: artificial life, machine learning, neural networks, robot learning, cognitive robotics, cognitive robotics lab, humanoid robots lab, humanoid robots seminar
computer-vision-audio: deep learning for visual recognition, audio signal processing
information systems: thesis in usable security and privacy, usable security and privacy, it security, sensor data fusion, mobile communication
only exercises: biomedical data science, humanoid robots, foundation of graphics, network security
Machine Learning in Python Datacamp
IBM Data Science in Python Coursera
Machine Learning A-Z in Python Udemy
Neural Networks and Deep Learning in Python Coursera
I worked in the campaign software unit within the CRM software of TEB bank for almost 1 year. Turkish Economy Bank is one of the top 10 largest and oldest banks in Turkey, headquartered in Istanbul.
During this period, I worked on tasks such as developing in-bank software with Java language and evam tool (in-bank software), updating existing codes, and designing applications for the needs of business units.
The Bilkent University is one of the top 5 research-based universities in Turkey, and it is located in Ankara. The THE/QS rankings are 500 in the world in 2018 to 2023.
The courses at the university highlight the combination of theory and practice, as well as algorithmic thinking, and the language of instruction is 100% English. During the university, I took many basic computer engineering courses as well as elective courses. The projects can be seen in my previous github account. The programming languages in classes are mostly Java and C++. The transcript and diploma supplements can be seen here.
main classes: final project: Helthscope App, software engineering project management, systems analysis and design, principles of engineering management, software product line engineering, artificial intelligence
compulsory classes: algorithms, programming, object-oriented software engineering, data structures, operating systems, database systems, calculus, linear algebra, probability and statistics
internships: System Administrator at Turkey Is Bank (1.5 month), Android Java Developer at Mia Technology (1 month)