Harun Yahya Ünal

Built an ASP.NET Core based competition server simulator for the TEKNOFEST Armed UAV Competition. The project implements REST API endpoints for UAV telemetry collection, QR/HSS coordinate services, lock-on & kamikaze reporting, and team authentication, using a layered architecture (Controller-Repository-DBContext) with DTO/AutoMapper patterns.

Managed command and telemetry streams using ROS 2's C# wrapper, designing a thread-based architecture to efficiently handle concurrent data flow. Stored our own UAV's and competitor UAVs' live and historical position data in PostgreSQL, powering the ground control station's live map visualization, historical route analysis, and tactical decision support. Structured the system into modular layers — ROS 2 integration, multi-threaded concurrent data processing, business logic, and database — covering both real-time communication and historical data querying in one unified architecture. Unfortunately our team was eliminated at the Critical Design Report (CDR) stage.

As software team captain, I built the software for a vehicle made of three units: an ESP32-S3, a Jetson Orin NX, and a ground control computer. I wrote the embedded firmware for the ESP32-S3. On the Jetson Orin NX, I built LiDAR-based SLAM mapping and a Nav2 navigation stack. For the ground control computer, I developed a React interface. All three units talked over ROS 2, except for the camera feed. The ESP32-S3 used the micro-ROS library with a micro-ROS Agent, so it could receive the ROS messages published by the navigation stack. The ground station used rosbridge to turn ROS messages into data the interface could read. The live camera feed went from the Jetson Orin NX to the ground station over WebRTC, and a FlySky FS-i6S RC transmitter handled manual control. Unfortunately our team was eliminated at the Critical Design Report (CDR) stage.

Developed a real-time target tracking and coordinate monitoring system (AegisTrack) within HAVELSAN's Surveillance and Subsystems Team. Built a rapid Python prototype first, with AI, GPS data processing, and H.264 video streaming modules, plus a React map interface for listing objects, drawing shapes, and displaying live coordinates. Based on the validated prototype, built the production version on Java Spring Boot with a layered architecture (Controller → Service → Repository), using PostgreSQL and PostGIS for geospatial storage, and Hibernate Spatial with JTS for CRUD on geometric data. Delivered live location updates via WebSocket, secured the API with Spring Security, and documented it with Swagger/OpenAPI — all containerized with Docker and built through Git/Pull-Request based team collaboration.

Published my personal website using a self-hosting approach on a Raspberry Pi, running the application in Docker and managing deployment through Dokploy. Exposed the site to the internet securely via Cloudflare Tunnel, without opening any ports directly, gaining hands-on entry-level experience with Linux server administration and network configuration.

Took on backend development and deployment for our mobile app, which uses React Native on the frontend, a microservice architecture on the backend, and an advanced RAG pipeline on the AI side. Since the app is feature-rich and each feature is its own domain, I built it as Java Spring Boot microservices behind an API Gateway that routes client requests to the right service, added a Celery/Redis pipeline that generates content on a schedule, and used three databases (PostgreSQL, MongoDB, Redis) for their respective purposes. We self-hosted the backend on our own Raspberry Pi server as well, brought up with Dokploy and exposed securely via Cloudflare Tunnel.