Mathematical intuition, applied to systems software

Roughly in chronological order, from an early fascination with spaceflight to the software I build today. The figures sketch the idea behind each chapter; most are computed from a small model of that idea rather than drawn by hand.

1. 01/Childhood – 2015

   Looking up: space exploration

   My connection to space began when I was very young, with the film Contact. After it, I could not stop reading and researching. The Moon missions made a deep impression on me, and I read everything I could find about them. Then came the Mars missions, and the probes that were lost or failed in space caught my interest just as much.

   I researched the space race between the Soviet Union and the United States from start to finish and watched all the films about it. In 2015 I started building small space-related projects of my own.

   One old project fascinated me in particular: a submarine for Titan. Titan is Saturn’s largest moon, and its surface has seas of liquid methane and ethane. The study, prepared for NASA in 2014–2015, proposed exploring the largest of these seas, Kraken Mare, with a submarine.

   The interest never left. It runs through the rocketry chapters below, and through an article I wrote in Turkish about the Apollo 11 crew’s visit to Anıtkabir in 1969.

2. 02/2016 – 20192016 to 2019·C# · MSSQL

   Programming foundations

   I started programming in 2016 at a vocational high school (meslek lisesi), in C#. The fundamentals came first, then data structures and algorithmic problem-solving, with small mathematical calculation engines along the way. By 2019 I was building complete database applications on Microsoft SQL Server (MSSQL).

   My graduation projects were desktop versions of tic-tac-toe and Battleship. Battleship was the game I played with my father as a child, which is why I wanted to design my own; the C# source of that first version is on GitHub. The Hunt & Target algorithm that now aims the computer’s Battleship shots came later: it hunts for ships across the board, then closes in on every hit.

   Both games now have browser versions on this site, with algorithmic opponents: Hunt & Target for Battleship, and a minimax search with alpha–beta pruning for tic-tac-toe. Play Battleship or tic-tac-toe.

3. 03/2019 · TÜBİTAK

   Research: cryptography and visual programming

   In 2019 I carried out a high-school research project with TÜBİTAK, the Scientific and Technological Research Council of Türkiye, on cryptography and the logic of visual programming.

   Part of it was an early prototype of a visual coding tool based on flowcharts: a program is put together by drawing its flowchart instead of typing it as text.

4. 04/2019 – 20222019 to 2022·Rocket and Space Club

   Rocketry and flight avionics

   I started university in November 2019 and joined the Istanbul University Rocket and Space Club, where I was Vice President until 2022. Our team designed and built three rockets: one low-altitude rocket with a 5,000 ft target altitude and two high-altitude rockets with a 10,000 ft target.

   I was solely responsible for the flight avionics firmware and the parachute deployment control system. Both relied on onboard sensor fusion of orientation, gyroscope and altimeter telemetry. How we got a good estimate out of that noisy data with a Kalman filter, and where we used PID controllers, is told inthe next chapter.

   I also designed the protocols for logging telemetry to an SD card, which worked alongside the RF telemetry transmission modules, and built a desktop telemetry dashboard that parses the flight data and renders post-flight trajectory plots.

5. 05/Rocket and Space Club · Kalman · PID

   From noisy data to a good estimate: Kalman filters and PID

   Data from a rocket’s sensors is never clean: the altimeter reads a slightly different value every moment, and the accelerometer picks up the motor’s vibration too. So while developing the rockets, we researched passing the measured data through a Kalman filter. We wrote the algorithms and tested them again and again.

   At every step a Kalman filter first predicts where the rocket should be, then corrects that prediction with the new reading. How much it trusts each one depends on how uncertain each is. Think of walking through a dark room counting your steps, and correcting your guess when you touch a wall; the filter does the same.

   In experiments such as landing a rocket vertically, we used PID controllers. A PID controller looks at the gap between the target and the measured value in three ways: how far off we are now (P), how long we have been off (I), and how fast the gap is changing (D). We do the same without thinking when we set the temperature in a shower.

6. 06/Research

   Autonomous parachute guidance and precision landing

   The autonomous parachute was a purely experimental and very enjoyable project I did with friends at university. The question was simple: once the parachute is open, can the payload steer itself to a chosen point on the ground?

   One of our friends solved the mechanical side. The software was mine: I wrote a decision algorithm built on linear algebra, and a simulation environment to try it in. A few of our friends then tested it very well.

   We read many earlier papers on the subject and learned a great deal from them. For the angle calculations we used the law of cosines and many trigonometric identities, and functions such as atan2 made the decisions robust: unlike a plain arctangent, atan2(Δy, Δx) knows which quadrant the angle lies in, and it still works when Δx is zero.

   The Python code is public on GitHub: AutonomousParachute.

7. 07/Python (2020) → C++ (today)

   Flight simulation: from prototype to Rocket-Up

   The flight simulation is one project that has changed shape over the years. It began in 2020 as a Python prototype: a three-dimensional numerical simulation of rocket trajectories, narrow and built for a single purpose.

   I am now rewriting it from scratch as Rocket-Up, a dynamic, open-source aerodynamic simulation engine in modern C++, built for higher performance and a clean architecture.

8. 08/Rocket and Space Club · CubeSat

   CubeSat: the satellite we could not finish

   Towards the end of the Rocket and Space Club, we decided to build a CubeSat, a small cube-shaped satellite. Ours was to follow every rule of the CubeSat standard. NASA had a PDF guide for CubeSat developers, and we learned a lot from it: how to solder, and how to design a CubeSat that is ready for launch.

   We did a lot of research. We thought through and designed the solar panels, the software that would run inside the satellite and even the ground station software. I was going to program the boards inside the satellite, and also write the ground station software: a dashboard running on a computer.

   But budget cuts and the club breaking up left the project half-finished. It was a really good project; we just could not finish it.

9. 09/2021 – 20242021 to 2024·crowd.inc

   Software Developer at crowd.inc

   crowd.inc was a website where people shared their ideas and found help for them. It had hundreds of users and was home to many projects, ideas and goals. Later we turned towards private ideas.

   From July 2021 to March 2024 I was a Software Developer there. I owned the software development lifecycle end to end: relational schema design in PostgreSQL, RESTful APIs with Python/Flask and jQuery, Linux server provisioning and automated testing.

   I designed granular role-based access control (RBAC) with strict boundaries between public and private data, built scalable feed pagination and dynamic data loading, and kept the platform reliable with rigorous unit and integration testing.

10. 10/GDSC · Mathematics Club

    Community: workshops, live streams and competitions

    In 2023 I was on the Google Developer Student Clubs (GDSC) core team. I hosted technical workshops and live streams on Flask, HTML and Git/GitHub, and organised Cyber Security Week, which included a live-streamed technical interview with a CCIE-certified network security engineer.

    In my later years at university I also organised and ran academic events for the Mathematics Club: seminars, and logic and mathematics competitions.

11. 11/Today

    Current core: modern C++ and Go

    Most of my time now goes into three kinds of software, mostly written in modern C++ and Go: system tools, telemetry and device-management software, and algorithm-driven applications. The projects page has the full list; two examples follow.

    Asion, which I started in 2024, is a privacy-first productivity and activity tracker for macOS, Linux and Windows, built with self-hosting in mind. It rests on native OS event hooks, a lightweight daemon architecture and gRPC between its components, and is in early access.

    Novacast began in 2025 at a computer company where I worked, as a replacement for software that was bloated and fell short of our needs. It manages many screens of different makes from one web panel: Windows, macOS and Linux machines, and Samsung, LG and Android TVs. A small Rust player runs on each screen, and a Go server reaches the screens through an MQTT broker. It is now in production.

    Alongside this I am completing my mathematics degree at Istanbul University. Real analysis, abstract algebra and topology rarely appear in my code directly, but the habits they teach, defining things precisely and checking every step, carry straight over.

12. 12/Today · CI/CD

    Automation: one-click test, package and deploy

    Shipping native software to three operating systems is only practical when the routine steps run themselves. I build multi-platform pipelines from GitHub Actions workflows and cross-compilation runners, tied together with my own scripts in Bash, Batch and Python.

    For the Asion ecosystem this became one-click pipelines: a single trigger tests the code, packages it, generates the artifacts for every platform and deploys them.