Open to ML / AI Engineering roles

Onur Altay Topaloğlu

Machine Learning Engineer & AI Researcher

Computer Engineering graduate from METU building trustworthy, multimodal AI systems — from medical vision-language models to large-scale computer vision research.

Istanbul, Türkiye M.E.T.U. — Computer Engineering
Get in touch Download CV

01. About

I'm a Computer Engineering graduate from Middle East Technical University (METU) with a CGPA of 3.65/4.00, focused on building machine learning systems that are reliable, fair, and ready for the real world.

My work spans applied research and shipped products: I design evaluation frameworks and retrieval systems, train and benchmark deep learning models, and turn research ideas into working pipelines — including ÜSTAT, an AI-native workspace I'm building for lawyers. I care most about the parts of ML that decide whether a system can actually be trusted in production — robustness, fairness, and rigorous evaluation.

Currently I'm a co-first author on a CVPR 2026 workshop paper and an undergraduate researcher in a Harvard–MIT × METU collaboration on trustworthy multimodal AI for healthcare.

Onur Altay Topaloğlu

Education

Middle East Technical University (METU)

Middle East Technical University (METU)

B.S. in Computer Engineering (English)

CGPA: 3.61 / 4.00

2021 – 2026

2021

University entrance exam

top 0.02% (544th of 2.4M)

Kadıköy Anatolian High School (Kadıköy Maarif College)

Kadıköy Anatolian High School (Kadıköy Maarif College)

High School Diploma (STEM)

2016 – 2021

2016

High-school entrance exam

top 0.7% nationwide

02. Research & Publications

CVPR 2026 — AI4Space WorkshopROMER, METUCo-First Author

LuMon: A Comprehensive Benchmark and Development Suite with Novel Datasets for Lunar Monocular Depth Estimation

A benchmark and development suite for lunar monocular depth estimation. Built an evaluation suite with affine alignment and dynamic masking for robust metric analysis, used Spearman rank correlations to quantify the impact of training scale and metric supervision on zero-shot domain adaptation, and engineered a ground-truth depth generation pipeline for real Chang'e-3 lunar imagery.

PaperProject PageGitHubHugging Face
In final revisions for submissionHarvard–MIT × METUCo-Author

Fairness Evaluation of Medical Vision-Language Models against Demographic Biases

Trustworthy multimodal AI for healthcare, with a focus on evaluating the fairness of medical vision-language models against demographic biases, and developing evaluation frameworks to analyze bias within medical imaging contexts.

03. Experience

Sep 2025 – Present

Remote

METUHarvard UniversityMIT

Undergraduate Researcher

Harvard–MIT & METU

Advisors: Dr. Jiaee Cheong, Prof. Dr. Sinan Kalkan

  • Co-authoring a paper (in final revisions) on the fairness of medical vision-language models against demographic biases.
  • Building evaluation frameworks to surface bias in medical imaging contexts.

Jul 2025 – Aug 2025

Istanbul, Türkiye

KUIS AI Center

Computer Vision Intern

KUIS AI Center, Koç University

Advisor: Asst. Prof. Fatma Güney

  • Studied flow matching and diffusion models for point tracking on objects with ambiguous motion.

Nov 2024 – Mar 2026

Ankara, Türkiye

ROMER, METU

Undergraduate Researcher

ROMER, METU

Advisors: Assoc. Prof. Gökberk Cinbiş, Prof. Dr. Sinan Kalkan

  • Co-first author of LuMon (CVPR 2026 AI4Space).
  • Built the evaluation suite — affine alignment, dynamic masking, and Spearman-based analysis of how training scale drives zero-shot transfer.
  • Engineered a ground-truth depth pipeline for real Chang'e-3 lunar imagery and released the datasets on Hugging Face.

Sep 2024 – Jun 2025

Ankara, Türkiye

METU

Undergraduate Teaching Assistant

Computer Engineering Department, METU

  • Led labs for CENG213 Data Structures (C++) and CENG232 Logic Design.

Jun 2024 – Jul 2024

Istanbul, Türkiye

KKB — Kredi Kayıt Bürosu

Artificial Intelligence Intern

KKB — Kredi Kayıt Bürosu

  • Built a hybrid RAG system (PyTorch) fusing vector and keyword search to improve relevance ranking over a keyword baseline.
  • Generated privacy-safe synthetic datasets with SDV for financial applications.

04. Projects

In development · Private

ÜSTAT

AI-native workspace for lawyers

From filing a case to the appeal deadline, ÜSTAT brings a lawyer's entire day into a single workspace — cases, clients, documents, and deadline tracking, all in one source of truth, paired with a document-aware AI assistant that understands the files it works with.

Legal TechAI AssistantRAGDocument Intelligence

Capture & Cook

Fridge photo → recipe suggestions · Senior Design Project

An app that turns a photo of your fridge into recipe suggestions. Built a search engine ranking recipes by fusing TF-IDF with dense semantic vector search using FAISS, optimized rankings via a custom cross-scoring algorithm trained on a human-preference dataset, and integrated the Alibaba Qwen VLM for automated ingredient detection.

FAISSTF-IDFSemantic SearchVLMPython
Website

05. What's next?

Let's build something together

I'm currently open to machine learning and AI engineering opportunities. Whether you have a role, a project or business idea, or just want to talk shop — my inbox is always open.

Say hello Download CV
altay.topaloglu@metu.edu.trGitHubLinkedInGoogle Scholar