Developer · Linux enthusiast · aspiring DevSecOps engineer
I’m a high-school student in Türkiye. Most of what I know came from building something, breaking it, reading until it made sense, and trying again.
01 / Projects
A small, local-first forensic and security device built around a Raspberry Pi. It explores network monitoring, audit logs, hash chains, integrity verification, RTC handling, mirrored storage, and WORM-style preservation.
The interesting part is not the box itself. It is figuring out how to collect information while making later tampering obvious.
A fine-tuning experiment based on LFM2.5-8B-A1B: roughly 109,000 examples, 2,000 supervised fine-tuning steps, and 800 preference-optimization steps. A practical look at what model training involves beyond calling an API.
02 / About
I’m Recai, a developer and Linux enthusiast following the IB Diploma Programme. My higher-level subjects are Mathematics AA, Physics, and English. Outside school I spend most of my time around backend development, infrastructure, networking, security, electronics, and open-source software.
My long-term direction is DevSecOps. I’m still learning—especially in penetration testing and security research—and I’d rather show the work honestly than pretend to be a senior engineer.
I like systems that can be opened, modified, rooted, reprogrammed, self-hosted, repaired, and understood. An old phone or laptop with an unlocked bootloader is usually more interesting to me than a polished device designed to stay closed.
03 / Working set
System administration, systemd, SSH, networking, logging, self-hosting, recovery, automation, and turning spare hardware into useful servers.
Software that talks to operating systems, networks, APIs, infrastructure, and hardware. I prefer practical tools over decorative abstractions.
Network reconnaissance, packet analysis, Wi-Fi and BLE, Android security, forensics, evidence integrity, and defensive automation.
ESP32, ESP8266, STM32, Raspberry Pi, RP2040, displays, radios, sensors, and inexpensive modules soldered into questionable prototypes.
ThinkPads, old Android phones, BlackBerry devices, handhelds, cyberdecks, custom ROMs, Termux, ADB, Fastboot, and alternative operating systems.
Ollama, quantized models, Hugging Face, fine-tuning, coding agents, and where AI helps—or becomes dangerous when nobody understands the generated code.
04 / Now
I’m currently improving my Linux, networking, backend, security, and electronics fundamentals while building projects that force those subjects to meet each other. The plan is simple: keep making things, document what worked, and be honest about what did not.
“AI should amplify technical knowledge, not replace the need to understand the system.”