Skip to content
Berat Kerem Öztopuz
Portrait of Berat Kerem Öztopuz

Computer Engineer · Ankara, Türkiye

Berat Kerem Öztopuz

ML, Backend and Game Development

I train models in PyTorch and put them into production with tested, reproducible pipelines and APIs. I also write iOS apps in SwiftUI and make 3D games in Unity.

GitHubLinkedIn

01Selected work

  • 2026

    Industrial Anomaly DetectionFeatured

    A model that finds defects on factory parts after seeing only good ones. It is PatchCore-style and trained on MVTec AD. Training and evaluation run as a Prefect pipeline, and the model is live as a FastAPI service on Google Cloud Run, where you can upload an image and try it.

    • Image AUROC 0.977 and pixel AUROC 0.977 across all 15 categories
    • Every run saves its source commit, dataset checksum and working-tree state, so each result can be traced back to the exact code
    • Packaged with Docker, tested in CI, and checked for data drift by a scheduled job on Cloud Run
    • PyTorch
    • Prefect
    • FastAPI
    • Docker
    • Cloud Run
    • GitHub Actions

    CodeLive API

  • 2026

    Skinner

    My graduation project. You take a photo of your face and a CoreML model on the phone analyses your skin. The app keeps your scores and routines, so you can see how things change over time. It is set up for App Store release with a StoreKit subscription.

    • Swift
    • SwiftUI
    • CoreML
    • Core Data
    • StoreKit

    Code

  • 2026

    Skin Condition Model

    The model inside Skinner. It checks a face photo for acne, eczema, eye bags and wrinkles separately, because one face can have more than one. It runs on ConvNeXt-Tiny, and each condition has its own threshold tuned on the validation set.

    • PyTorch
    • ConvNeXt-Tiny
    • CoreML

    Code

  • 2026

    Pneumonia Detector

    Looks at a chest X-ray and tells you whether it shows pneumonia. I fine-tuned ConvNeXt for this and added Grad-CAM, which marks the parts of the X-ray the model paid attention to.

    • PyTorch
    • ConvNeXt
    • Focal Loss
    • Grad-CAM

    Code

  • 2025

    Art Movement Classification

    Give it a painting and it guesses the art movement: Baroque, Cubism, Pop Art or one of seven others. Transfer learning on ConvNeXt-Tiny got it to about 75% accuracy. This was my term project for the Deep Learning course.

    • PyTorch
    • ConvNeXt-Tiny
    • CutMix
    • Focal Loss

    Code

  • 2026

    The House Always Wins

    A casino game where you run security and catch cheaters. Six of us made it, and I wrote all of the NPC AI. Customers walk the floor with NavMesh and move between blackjack, roulette, slots and the bar. Some of them cheat, some only look suspicious, and you decide who goes to the interrogation room.

    • Unity 6
    • C#
    • NavMesh
    • State machine

    Code

  • 2026

    Core Breach

    A first-person game in a sci-fi arena. You hold off enemies until the timer runs out. Some go for the energy core and some come after you. In the code I used Strategy for enemy behaviour, Observer for the core's health, an object pool for projectiles and decorators for weapon upgrades.

    • Unity 6
    • C#

    Code

  • 2026

    Flight Mission Simulator

    Take off, get through a danger zone without being hit by homing missiles, then land. The missiles chase the plane using direction vectors and quaternion rotation. A mission manager handles the four game states and the HUD.

    • Unity 6
    • C#
    • TextMesh Pro

    Code

  • 2026

    Endless Flight Obstacle Game

    Fly as far as you can between towers that keep appearing in front of you. You control throttle, pitch, roll and yaw, and a Cinemachine camera follows the plane. Towers you have passed get removed, so the scene does not fill up.

    • Unity 6
    • C#
    • Cinemachine

    Code

  • 2026

    Gym Management System

    Term project for the Database course. It is a Flask site on PostgreSQL where members sign up, book group lessons and track their BMI. A single setup script creates the schema and fills it with sample data.

    • Python
    • Flask
    • PostgreSQL

    Code

  • 2023

    Temperature & Humidity Controller

    Built during my internship at Leonardo Türkiye. I wrote the frontend.

    • JavaScript
    • HTML
    • CSS

    Code

02Experience

  1. 2026

    Wraith Esports · Intern

    • Two of us built Wraith's storefront app from start to finish.
    • I designed the UX, then turned it into a design system of reusable components that the whole app uses.
  2. 2023

    Leonardo Türkiye · Intern Engineer

    • Three of us built a system that monitors temperature and humidity in real time.
    • I wrote Flask REST endpoints that validate incoming requests, and the JavaScript frontend that shows the sensor data.

03Skills

Languages
Python, C#, Swift, JavaScript, Java
ML & AI
PyTorch, MLflow, ConvNeXt, PatchCore, Grad-CAM, CoreML, CUDA
Backend & Data
FastAPI, Flask, REST APIs, PostgreSQL, Postman
Cloud & DevOps
Docker, Google Cloud Run, GitHub Actions, Prefect, Linux, Automated testing
Game Development
Unity 6, Rigidbody physics, Cinemachine, TextMesh Pro, Game state management
iOS
SwiftUI, Core Data, StoreKit, Xcode
Tools
Git, GitHub, macOS

04Education

2021-26

B.Sc. in Computer Engineering

University of Turkish Aeronautical Association

Game Programming, Deep Learning, Database Systems

05Contact

I'm looking for a junior role and I'm open to freelance work. The quickest way to reach me is LinkedIn.

LinkedInGitHub