Utku Ayaz Akkas

Utku Ayaz Akkas

M.Sc. Researcher · Embodied AI & Robot Learning

I'm an M.Sc. student at CogRob Lab, Özyeğin University, advised by Prof. Erhan Öztop. My research focuses on embodied and developmental learning — specifically on how pretrained vision-language models can act as priors that structure an agent's exploration, rather than only serving as perception modules. I'm also interested in open-ended novelty-driven discovery in robotic manipulation.

Istanbul, Turkey
M.Sc. Student, Ozyegin University
utkuayazakkas@hotmail.com
Contact Me

Publications

Robot Scaffolding with Vision Language Models

IEEE International Conference on Development and Learning 2026

2026

This paper proposes a scaffolding approach for robotic systems to explore the environment using vision language models.

Project Page

Tamper-Proof Evidence via Blockchain for Autonomous Vehicle Accident Monitoring

IEEE 1st Global Emerging Technology Blockchain Forum: Blockchain & Beyond (iGETblockchain)

2022

This paper proposes a blockchain-based solution for securing and validating evidence from autonomous vehicle accidents, ensuring tamper-proof records.

View Publication Cite Paper

Teaching Assistantships

Fall 2025

Courses:

CS 101 — Computer Programming
  • Conducted laboratory sessions to assist students with material
  • Grading
CS 201 — Data Structures
  • Grading and office hours

Professional Experience

Sep. 2025 – Present

Graduate Researcher

CogRob Lab, Özyeğin University — Advisor: Prof. Erhan Öztop

Istanbul, TR

  • Developed a vision-language scaffolding framework that proposes visually grounded pick-and-place actions for a simulated UR10e manipulator, treating a VLM as an exploration prior rather than a perception front-end.
  • Showed the method outperforms random exploration at reaching hard-to-reach configurations, and improves novel-state discovery in three of four evaluated settings.
  • Designed the evaluation protocol and ablations in PyBullet.
Apr. 2025 – Dec. 2025

Co-Founder & Software Engineer

PlayAction (sports analytics)

Istanbul, TR

  • Built a computer-vision web platform end to end — data model, backend, OpenCV analysis pipeline, and deployment — for automated analysis of match footage and performance data.
Sep. 2023 – Sep. 2024

Data Scientist

Borusan CAT

Istanbul, TR

  • Built predictive-maintenance models for heavy-equipment fault detection from operational telemetry.
  • Replaced legacy SAS forecasting workflows with a Python time-series pipeline.
  • Developed internal generative-AI tooling on Azure for automated document processing.
Aug. 2022 – Jul. 2023

Undergraduate Research Assistant

Özyeğin University — Advisor: Asst. Prof. Mehmet Parlak

Istanbul, TR

  • Coordinated the "Automatic Smart Contract Generation" project, including the evaluation pipeline for automatically generated Solidity contracts.
  • Designed and executed test suites assessing correctness and reliability of generated code; contributed to the blockchain evidence-integrity work published at iGETblockchain 2022.
  • Onboarded and mentored incoming research assistants on the codebase and experimental procedures.