Curriculum Vitae

Full Stack Developer Intern

Hepsiburada, Istanbul

07/2025 - 08/2026

• Developed and maintained internal backend services using .NET 10 and C#, implementing new REST API endpoints and improving existing ones.
• Designed and built an internal logging and observability service from the ground up using Kotlin and Spring Boot.
• Built and maintained micro-frontend (MFE) applications for the customer management team using React, TypeScript, and Tailwind CSS, with Redux for shared state management.
• Developed customer-facing pages for Hepsikredi (consumer credit) and Kobi Kredileri (SME lending), along with internal backoffice interfaces for the fraud and Hepsikredi teams.
• Refactored backend code to resolve production issues and improve system reliability across production and test environments.
• Configured and maintained CI/CD pipelines, containerized and deployed services to Kubernetes with Docker.
• Monitored application logs and system health using OpenSearch and Elasticsearch, with Slack alert integrations for incident response.

IT Intern

Türk Telekom, Tokat

09/2018 - 06/2019

• Assisted in troubleshooting hardware and software issues, supported system configuration and user environment setup.
• Gained exposure to enterprise IT infrastructure and operational workflows.

Project & Research Paper: Game Analysis with Image Processing

• Architected an end-to-end computer vision system that detects physical game tiles and recommends optimal moves, using a client-direct architecture with Next.js and an isolated Flask API to bypass serverless cold starts.
• Implemented a two-stage hybrid AI pipeline with YOLOv8-Nano for localization (99.4% mAP50, 41ms) and ResNet-18 for classification (97.26% accuracy, 2.88ms).
• Formulated complex combinatorial game rules with Depth-First Search and memoization (hash caching), calculating optimal 101-point and pair-opening strategies in under 10 milliseconds in-browser.
• Engineered a custom dataset of 4,432 labeled tiles across 55 classes, with a hybrid annotation pipeline in OpenCV and Python that cut manual labeling from 50+ hours to roughly 1 hour.
• Built an interactive React frontend for real-time manual correction of misclassified tiles, triggering instant algorithm recalculation with 150-250ms end-to-end latency.

Movie Recommendation Web App

• Built a full-stack recommendation platform on the MovieLens ml-32m dataset (32M ratings, 87,585 movies), with Google sign-in to rate movies, build a watchlist, and browse the catalog with genre/decade/rating filters.
• Implemented an offline-trained item-item collaborative filtering engine (asymmetric cosine similarity with MMR re-ranking for diversity) served client-side from a compact static neighbour table, with no MovieLens user data ever reaching the browser.
• Built with Next.js (App Router) and TypeScript, storing per-user ratings and watchlist data in Cloudflare D1 via Auth.js's Google adapter, deployed to Cloudflare Workers through OpenNext.

To-Do App

• Built a full-featured to-do app with a REST API and React SPA served from a single Cloudflare Worker (one origin, no CORS, one deploy), with Google OAuth sign-in, tags, subtasks, and drag-and-drop reordering.
• Implemented soft-delete with a trash view and undo, using Hono, Cloudflare D1 (SQLite) via Drizzle ORM, and Zod for request validation on the backend, with React 19, TypeScript, Vite, and Tailwind CSS 4 on the frontend.

Personal Portfolio Website

• Responsive Next.js and React portfolio with accessibility optimizations, deployed on Cloudflare Pages with automated builds.

Healthcare Sentiment Analysis with NLP

• Co-authored an NLP research paper published in “Yapay Zeka Tabanlı Sistemler: Teori, Uygulama ve Gelecek Perspektifleri-2” (BIDGE Yayınları), processing 12,600+ patient reviews from NHS and RateMDs APIs.
• Achieved classification accuracy of 76% F1-Score utilizing RoBERTa.

Analysis of Suicide Content in Social Media Posts with Deep Learning Models

• Authored a research paper on detecting suicidal intent in social media posts using deep learning models, implementing and comparing two novel models in this domain and achieving 97% accuracy.
• Conducted data preprocessing, feature extraction, and model evaluation to improve classification performance.

BSc in Computer Engineering

Marmara University, Istanbul

2022 - 2026