İsmail Ayberk Full Stack & AI Developer
TR

Hi, I'm İsmail Ayberk.

I work as a Software Developer at Doğuş Teknoloji. Alongside that, I work as an AI Developer on various projects.

Scroll down, or click the marked objects in the room.

about.txt

About

I study Computer Engineering at Konya Technical University and work as a Full Stack Developer at Doğuş Teknoloji.

Outside of work, I build AI projects. At Decyra, which I co-founded, I work on synthetic personas and LLM-based decision simulation. For Translation Feedback App, I designed and shipped the architecture of a system that generates translation feedback with a local LLM and explainable AI. At Aila, I work on the backend, embedded and hardware side of an assistant robot that can be customized for each business.

I'm vice president of the KTÜN AI & Data Science Lab, where I coordinate teams and teach Python and computer vision to new members.

location
Konya, Türkiye
now
Doğuş Teknoloji · Decyra
education
Konya Technical University, Computer Engineering
languages
Turkish, English
experience.log

Experience

  1. NOW Present

    Full Stack Developer

    Doğuş Teknoloji

    Working on backend and frontend development of enterprise applications.

    • ASP.NET
    • C#
    • Full Stack
  2. NOW May 2026 — Present

    Co-founder & Systems Engineer

    Decyra

    Building an AI-powered decision simulation platform that lets organizations test likely human reactions to a decision with synthetic personas before putting it into practice.

    • LLM
    • Synthetic Personas
    • Behavioral Modeling
    • Decision Simulation
  3. Jan 2024 — Dec 2025

    Project Developer / Research Fellow

    Konya Technical University × İMAŞ Makina — TÜBİTAK 1711

    Built an AI-based monitoring and predictive maintenance system for industrial band-saw machines, plus a Qt-based operator interface.

    • Python
    • Object Detection
    • Image Classification
    • PyQt
  4. NOW 2023 — Present

    Vice President · Project Developer

    KTÜN AI & Data Science Lab

    Coordinating teams, sponsorships and technical events; developing image processing projects.

    • Computer Vision
    • Mentoring
    • TÜBİTAK
    • TEKNOFEST
projects/

Projects

Ongoing May 2026 — Present

Decyra

Co-founder & Systems Engineer

Problem: Organizations often only see how employees or customers react to a high-impact decision after it has been rolled out.

Solution: We are building a decision simulation platform that tests decisions on synthetic personas before they are put into practice. I work on persona modeling, quantifying behavioral parameters and LLM-based simulation.

  • LLM
  • Synthetic Personas
  • Behavioral Modeling
  • Decision Simulation

Ongoing Oct 2026 — Present

Translation Feedback App

System Architecture · Full Stack · Deployment

Problem: Translation and interpreting students had no tool for getting quick feedback on their translations with the reasoning explained.

Solution: We built a feedback platform powered by an LLM and explainable AI (XAI). I designed the system architecture and built the Next.js web app, the database and job queue on Supabase, and a Python worker that runs a local Qwen model. The worker processes queued translations and produces LLM feedback and counterfactual explanations. I also handled server setup, deployment and the production launch.

  • Next.js
  • TypeScript
  • Supabase
  • PostgreSQL
  • Python
  • Qwen
  • XAI
  • Vercel

Ongoing 2025 — Present

Ailab Super App

Lead Backend Architect

Problem: Lab access, task tracking and reporting were scattered and manual.

Solution: I designed from scratch the backend of a lab management system with ERP and IoT integration, using .NET 8 and Clean Architecture. I built JWT + RBAC authorization, RFID entry/exit tracking wired to real devices, and task assignment and central reporting modules. The system serves 100+ active users.

  • .NET 8
  • ASP.NET Core
  • PostgreSQL
  • Clean Architecture
  • JWT · RBAC
  • IoT · RFID
  • Docker
  • Firebase

Completed Jan 2024 — Dec 2025

AI-Based Traceability, Efficiency and Predictive Maintenance for Next-Generation Band Saw Metal Cutting Machines

TÜBİTAK 1711 · İMAŞ Makina

Problem: Crack and wear inspection on industrial band saws was done by hand.

Solution: I built a hybrid vision pipeline combining object detection and classification, reaching over 98% accuracy in crack and wear detection. I tested it on real machines in the field and built a PyQt interface for operators.

  • Python
  • Object Detection
  • Classification
  • PyQt
  • Industrial AI

Ongoing 2025 — Present

Aila

Backend · Embedded Systems · Hardware

Problem: Digital screens in physical venues such as restaurants, hospitals or event spaces can't really interact with customers.

Solution: We are building an LLM-powered assistant robot that talks in natural language and can be customized with each business's own data. I work on the backend and embedded side; I designed the robot's entire outer shell in 3D and worked on the IoT components and electrical setup.

  • LLM
  • Backend
  • Embedded Systems
  • IoT
  • 3D Modeling
  • Robotics

Completed Apr 2026 — Jun 2026

KTÜN Computer Engineering Chatbot

RAG · Dataset · Full Stack

Problem: Students' questions about courses, internships, transfers and Erasmus were answered across scattered regulations and pages.

Solution: We built a Turkish chatbot that runs fully locally, with no internet or cloud, and compared different RAG architectures on the same dataset using RAGAS metrics. I worked on the RAG architectures, helped prepare the dataset and built the backend and frontend of the app.

  • Python
  • RAG
  • LangGraph
  • Qwen
  • RAGAS
  • Embeddings

Completed Dec 2023 — 2026

TEKNOFEST AI in Aviation

Finalist · 2nd Best Software (2024)

Problem: Detecting small, distant objects in aircraft camera footage and estimating position when GPS is unavailable.

Solution: I developed detection models with YOLO + SAHI and RT-DETR, and position/orientation estimation with ORB-SLAM and regression models. We reached 92% localization accuracy on recorded footage and were named second best software in the 2024 finals.

  • YOLO
  • RT-DETR
  • SAHI
  • ORB-SLAM
  • Regression

Completed Nov 2024 — Jun 2025

TEKNOFEST Fighter UAV

Vision & Tracking

Problem: Tracking an opponent UAV in real time without losing it during sharp maneuvers.

Solution: I combined object detection with a pose-estimation-based tracking algorithm. The system was built to run on real flight, simulation and recorded footage, and inference pipeline optimizations raised the FPS.

  • YOLO
  • Pose Estimation
  • Tracking
  • Real-Time Inference

Some of the source code is private due to institutional/project confidentiality. Happy to talk through the details.

inventory

Skills

Backend & Web

  • .NET 8
  • ASP.NET Core
  • C#
  • EF Core
  • Clean Architecture
  • REST API
  • Next.js
  • TypeScript
  • JWT · RBAC

AI

  • Python
  • LLM
  • RAG
  • LangGraph
  • RAGAS
  • XAI
  • YOLO
  • RT-DETR
  • Pose Estimation
  • SLAM

Data & Cloud

  • PostgreSQL
  • Supabase
  • SQL
  • Firebase

DevOps & Hardware

  • Docker
  • Git
  • Vercel
  • Swagger / OpenAPI
  • Qt / PyQt
  • IoT
  • RFID
  • 3D Modeling
contact.sh

Contact

For a project, a collaboration or a question, you can reach me by email.