Engineering Philosophy

Deterministic Decision Making in Non-Deterministic Environments

The real world is noisy, unstable, and full of hidden state. Most of my work explores how decision-making systems can remain repeatable, explainable, and auditable anyway.

My projects may look different on the surface — prompt tooling, LLM agents, game agents, edge AI, data collectors, desktop tools — but I keep returning to the same reliability problem: how to make decisions that can be inspected, reproduced, and trusted under real-world uncertainty.

The central thesis

The model can be correct. The system can still fail.

For a long time, I followed the standard recipe: more data, larger models, better metrics. Real systems challenged that belief.

The root cause is often not the loss function, the optimizer, or the architecture. It is hidden state, implicit dependencies, untracked changes, weak observability, and timing effects.

That is why I treat decision-making as a systems problem, not only as a model problem. I care more about understanding failure modes than showcasing demos.

Determinism is not a property of the model alone. It is a property of the surrounding system design.

Same Problem, Different Layers

Different projects, the same reliability question seen from different parts of the stack.

Prompt / Decision Logic

PromptLedger

If prompts are not versioned like code, decision logic becomes unreliable.

AI Workflow Reliability

PromptLedger, Local LLM Tools, Debugging Utilities

Before an AI workflow can be trusted, its prompts, inputs, outputs, and changes need to be inspectable.

Perception → Action

JumpNet

A correct decision made at the wrong time is still wrong.

Hardware / Edge

Pico Trend Alarm, Sound Classifier, Mini SCADA

Noisy sensors and limited resources force explicit state machines, hysteresis, and controlled behavior.

Data Control

Custom Collectors, Logging Tools, Dataset Viewers

Uncontrolled data produces unstable system behavior.

How I Work

The practical habits behind the philosophy.

Make decision logic inspectable

Prompts, rules, thresholds, workflow decisions, and fallback behavior should be traceable and versioned.

Control the inputs

Data collection, logging, and synchronization are part of the system, not side work.

Measure timing

In real systems, when a decision is made can matter as much as what decision is made.

Prefer observable systems

A system that cannot explain its behavior after failure is not reliable enough.

Build under constraints

Edge devices, latency budgets, and noisy inputs expose assumptions that clean notebooks hide.

Document the process

Mistakes, design decisions, and failure modes are part of the project, not things to hide.

Background Snapshot

Brief context, after the thinking that drives the work.

AcademicComputer Engineering Student @ RWTH Aachen University

ResearchWerkstudent Researcher @ Fraunhofer IAIS

LocationAachen, Germany

FocusAI workflow reliability, local AI, edge AI

Public ToolsOpenAnima, MediaRecorder Lite, PromptLedger

Experience & Credentials

A compact view of the work, learning, and mentoring behind the projects.

Experience Timeline

Fraunhofer IAIS

Werkstudent Researcher

Dec 2025 – Present Sankt Augustin, Germany · Hybrid

Working on applied AI research and prototype development for intelligent enterprise workflows. My work includes refining research prototypes, frontend and UI work, experimentation, technical analysis, and turning research ideas into usable prototype systems.

Applied AIResearch PrototypesFrontend RefinementTechnical AnalysisEnterprise Workflows

Q-GEN

Co-Founder

Mar 2023 – Dec 2023 Ankara, Türkiye · Remote

Co-founded an EdTech project focused on AI-assisted learning tools and personalized education workflows. I led a team of five across product planning, prototype development, and early platform design. The experience clarified the gap between an idea and a usable learning platform.

EdTechProduct DevelopmentAI-assisted LearningTeam CoordinationPrototype Development

TURK AI

Software Developer

Jun 2022 – Aug 2022 Ankara, Türkiye · On-site

Developed Edge AI and computer vision solutions using OpenMV, ESP32, and embedded systems, with a focus on constrained-hardware optimization, IoT integration, and real-time AI applications.

Edge AIComputer VisionESP32OpenMVEmbedded SystemsIoT

Selected Certifications

MathWorks

Machine Learning Onramp

MathWorks

Deep Learning Onramp

IBM SkillsBuild

Data Fundamentals

Saylor University

CS107: C++ Programming

Mentoring & STEM Outreach

Bilim Kahramanları Derneği

STEM Mentoring & Robotics Outreach

2020 – 2022

Mentored high-school robotics teams in FIRST LEGO League and World Robot Olympiad contexts, supporting students with robotics, problem solving, teamwork, and early engineering thinking.

COSMOS FIRST LEGO League
SAVİORS FIRST LEGO League
PROCYON_50_2 World Robot Olympiad

STEM MentoringRoboticsFLLWROTeam Coaching

Where This Shows Up

The deeper implementations and notes behind the philosophy.

Project Systems See PromptLedger, JumpNet, edge AI deployments, and public debugging utilities. Technical Writing Read the mistakes, design decisions, project write-ups, and failure modes. Lab Notes Explore smaller tools, prototypes, and experiments supporting the larger systems.

Outside the Main Work

The same curiosity, just in less formal settings.

Building & Experiments

Outside of coursework and research, I still tend to orbit around systems: small electronics projects, microcontrollers, desktop tools, and experiments that start as rough ideas and slowly become usable. I like the moment where software stops being abstract and has to deal with hardware, timing, interfaces, or unpredictable human behavior.

Movement & Timing

Away from the keyboard, I enjoy football and volleyball. They are different from engineering, but I like the same thing there too: fast decisions, timing, feedback, and adapting to a changing environment.

Explore More

Continue into the projects, notes, and public profiles.

ProjectsCurated systems and products WritingTechnical notes and project logs LabExperiments and prototypes GitHubSource code and repositories LinkedInProfessional profile and updates