about
Born 1991 in Diyarbakır. Trained as a mechanical engineer. Working as a senior test automation engineer. Building a self-hosted AI called Alcyone on a Raspberry Pi 5 after hours.
story
I was seven when I watched Back to the Future and decided I would build things. The idea stayed. Two years later we had an Atari clone — NES-class hardware that arrived in our part of the world a decade after it was current. A few years after that we had a real computer with no Windows on it, and for a few months I lived in DOS, typing commands into a black screen and watching something happen. That was where I switched ambitions: physical-world science felt finished; the digital frontier was still empty. From that DOS prompt forward I wanted to be a computer engineer.
I failed the university entrance exam the first year and didn't place into computer engineering; tried it again the next year and placed into mechanical engineering at Istanbul University. The bachelor's took six years instead of four; the thesis was on the position control of linear motors — LabVIEW driving a National Instruments DAQ card, an Arduino generating the PWM signals, a stepper-style control loop wrapping it all. The mechatronics master's at Marmara that followed made software inevitable. Coursework took me through MEMS, ergonomics, control theory, and a long run of machine-learning projects. One of them — a CNN built on VGG16 transfer learning — hit 98.3% test accuracy on a 40,000-image cement-surface defect dataset. The thesis itself is on motor predictive maintenance and is still waiting for its defense date.
The software career began earlier than the diploma. The first repo on my GitHub is from September 2017, and there are forty-something more behind it — some graveyards, some foundations, none deleted. Five years on telecom-grade systems taught me how to ship at scale; the last two years on test automation taught me how to ship cleanly. In January 2024, well before the word agentic was the entire industry, I built dAIlog: a multi-agent framework on top of AutoGen. A single sentence to the system would route through seven agents — a Planner, an Engineer who wrote Python, an Executor who ran it, a Critic who checked the work — and something would come out the other end. I proposed it internally. It went nowhere. Six months later everyone caught up.
What is on the bench right now: a self-hosted AI on a Pi called Alcyone; Thalamus, a community plugin for it; a 9-layer testnet trading bot that runs continuously on the same Pi; an AI QA agent that writes its own end-to-end specs; and Ember, an RPG built by hand in Unity after the AI-scaffolded prototype proved unplayable. Underneath them is the thread you can read from 2017 onwards: ship the artefact, keep going. An Estonian e-Residency application is in flight.
experience
Test automation services firm
Global telecom equipment company
education
Marmara University
Thesis: "Predictive Maintenance for Electrical Motors by Using Machine Learning." Coursework done, manuscript on the desk, defense pending. A second master's project — image-processing-based defect detection — reached 98.3% test accuracy on a 40,000-image cement-surface dataset using VGG16 transfer learning.
Istanbul University
Thesis on the position control of linear motors — LabVIEW driving an NI PCI-6289 DAQ, Arduino producing the PWM signals. Six years of CAD, control theory, and the slow physics-first habit of mind that has stayed with me since.
stack
Languages
Test automation
AI / ML
Backend
Infrastructure
Hardware
Game / engine
also