AI systems that see, measure, and decide.

AI/ML engineer. Retrieval pipelines and LLM agents that show their evidence; robotic inspection cells that measure to spec.

case: withdrawal not creditedPLAN/tx/lookup/kyc/status/wallet/ledgerevidenceEVALUATEtenant scope isolatedstatus enums groundedevidence provenanceANSWER, with receipts
The investigation agent: plan, retrieve, verify, then answer.

Selected work

Jaguar

Robotic quality control, from laser scan to digital twin

An automated inspection cell I led end to end at MND Robotics: multi-angle laser scans are reconstructed into dense 3D point clouds, registered into a digital twin, and measured against spec in about 40 seconds per part.

A* path planning drives the scanner around complex geometries, while a React and Flask dashboard gives operators live control, monitoring and integration with external QC equipment.

scan to report, per part
~40s
scan to report, per part
digital twins of scanned parts
3D
digital twins of scanned parts
autonomous scanner path planning
A*
autonomous scanner path planning
  • ROS 2
  • Open3D
  • OpenCV
  • Mech-Eye
  • Docker
  • Linux
  • React
  • Flask
  • Sentry
Open live demo
Laser sweep reconstructing a measured part.

Projects

AI Investigation Agent

Consolidated support and finance escalations into a Zendesk hub, cutting resolution time by 40%. Then layered an AI investigation agent on top: it plans, executes and evaluates, pulling evidence through hybrid dense/sparse retrieval and multi-LLM routing across 76 backend endpoints and 29 categories, and gates its own conclusions behind an evidence checklist so it cannot answer before it can show its work.

A case that cost the dev team two developer-days closed in minutes, for about $0.03 in model spend

  • Chroma
  • BM25
  • RRF
  • Ollama
  • OpenRouter
  • FastAPI

AGV Migration for TAI

Migrated an autonomous guided vehicle delivered to Turkish Aerospace Industries from ROS 1 to ROS 2, and kept its software healthy across multiple customer sites.

Modernized autonomy stack in production

  • ROS 1
  • ROS 2
  • Linux

AI Influencer Platform

Persona-driven content engine that ingests news via RSS and RAG, then drafts and schedules platform-specific posts for marketing teams.

Scheduled multi-platform content on autopilot

  • RAG
  • LLM personas
  • RSS

EEG Ocular Artifact Denoising

B.Sc. thesis: a U-Net deep learning pipeline that removes EOG artifacts from 19-channel EEG signals, clean enough for downstream analysis.

Deep learning for biomedical signal processing

  • U-Net
  • TensorFlow/Keras
  • SciPy

Experience

From robotics R&D and manufacturing systems to applied AI.

  1. Nov 2025 - Aug 2026

    AI/ML Engineer

    Crypto Exchange, Tirana, Albania

    • Rebuilt the support and finance escalation workflow around a centralized Zendesk hub, cutting average resolution time by 40%.
    • Extended it with an AI investigation agent using hybrid retrieval and multi-LLM routing across 76 endpoints and 29 categories, cutting resolution time a further 80% and closing a case that had cost the dev team two developer-days in minutes, for about $0.03 in model spend.
    • Shipped an AI influencer platform generating scheduled, persona-driven content from RAG and RSS ingestion.

    Chroma, BM25, RRF, Ollama, OpenRouter, Zendesk, Python, FastAPI

  2. Aug 2024 - Nov 2025

    R&D Engineer

    MND Robotics, Antalya, Turkey

    • Led the end-to-end development of Jaguar, an automated quality-control system combining multi-angle laser scanning, point-cloud reconstruction and A* path planning into 3D digital twins in roughly 40 seconds per part.
    • Built the React-Flask operator dashboard for real-time control and monitoring, integrated with external QC equipment.
    • Migrated a Turkish Aerospace Industries AGV from ROS 1 to ROS 2 and supported deployments across customer sites.

    ROS 2, Open3D, OpenCV, Mech-Eye, Docker, React, Flask, Sentry

  3. Aug 2023 - Sep 2023

    R&D Intern

    IMECAR, Antalya, Turkey

    • Prototyped ROS 2-based software for an autonomous waste-collection robot.

    ROS 2, Python

  4. Jun 2023 - Aug 2023

    Full Stack Developer Intern

    DT Cloud, Antalya, Turkey

    • Built and deployed a full-stack ML classification pipeline: React frontend, Flask backend, a PostgreSQL data layer, and JWT-based authentication and session handling.

    React, Flask, PostgreSQL, JWT

Skills

AI & Machine Learning

  • TensorFlow/Keras
  • PyTorch
  • LoRA
  • vLLM
  • U-Net

Retrieval & Agents

  • Hybrid retrieval
  • BM25
  • RRF
  • ChromaDB
  • RAG
  • Tool calling
  • Structured outputs

Robotics & 3D

  • ROS 2
  • Open3D
  • OpenCV
  • Point-cloud processing

Engineering

  • Python
  • FastAPI
  • Flask
  • React
  • Docker
  • Linux
  • PostgreSQL
  • Redis
Turkish
Native
English
Professional working proficiency
OPEN TO ROLES

Hiring for AI/ML or robotics?

I am actively looking for my next role: agent systems, retrieval, or robotics. Happy to talk through any of the work above, or whatever you are building.

© 2026 Eyuphan Durmus

hey@eyuphan.devBack to top