# Mert Kilickaya

> Senior AI Research Engineer. Senior Research Engineer at Greyparrot.

I am an AI researcher. My goal is to help automate vision. Can we build models that curate their own data, extract their own supervision, and adapt on the fly to changes in the real world?

I want visual models to take a more active role in their own learning: choosing useful data, learning from new observations, and adapting efficiently as conditions change. This goal connects my work in self-supervised, continual, and vision-language learning with applications from medical diagnostics to industrial visual intelligence.

Canonical website: https://kilickaya.github.io/

## Education

I completed my PhD in computer vision at the University of Amsterdam’s Qualcomm Labs, supervised by Arnold Smeulders. Since then, I’ve been working in industry as a research engineer.

- Lab: [Qualcomm Labs](https://ivi.fnwi.uva.nl/quva/)

- PhD supervisor: [Arnold Smeulders](https://scholar.google.com/citations?user=aa5Ou7gAAAAJ&hl=en)

## Profiles & contact

- [Email](mailto:kilickayamert@gmail.com)

- [CV](https://kilickaya.github.io/resume/resume.pdf)

- [Google Scholar](https://scholar.google.com/citations?user=MBMjO0sAAAAJ&hl=en&oi=ao)

- [LinkedIn](https://www.linkedin.com/in/mert-kilickaya/)

## Experience

### Greyparrot

Senior Research Engineer · October 2026 — present

Research focus: Industrial visual intelligence

### Agendia

ML Researcher · 2023 — October 2026

Research focus: Biomarker Discovery

Location: Amsterdam, Netherlands

### Eindhoven University of Technology

ML Researcher · 2022 — 2023

Research focus: Continual Pretraining Agents

Location: Netherlands

- [Learning to Learn Lab](https://dai.win.tue.nl/team/5/)

### Huawei Visual Search Lab

Research Intern · March 2021 — January 2022

Research focus: Multimodal Learning

Location: Helsinki, Finland

### Qualcomm Labs, University of Amsterdam

PhD Researcher · 2017 — 2022

Research focus: Deep Learning for Vision

Location: Netherlands

- [Qualcomm Labs](https://ivi.fnwi.uva.nl/quva/)

- [Arnold Smeulders](https://scholar.google.com/citations?user=aa5Ou7gAAAAJ&hl=en)

## News

- **2026-10** — Our work at Agendia on AI-driven digital pathology for the 70-gene recurrence signature was presented at ESMO 2026, the most prestigious conference in Oncology.

  - [AI-driven digital pathology for the 70-gene recurrence signature](https://agendia.com/agendia-ai-driven-digital-pathology-esmo-2026/)

- **2026-10** — I joined Greyparrot as a Senior Research Engineer.

- **2024-10** — Honored to be an outstanding reviewer for ECCV'24!

  - [Honored to be an outstanding reviewer for ECCV'24!](https://kilickaya.github.io/resume/eccv_reviewer.pdf)

- **2024-07** — Our work on Transformer Architecture Search is accepted at ECCV'24!

  - [Our work on Transformer Architecture Search is accepted at ECCV'24!](https://arxiv.org/pdf/2407.16269)

- **2023-09** — Our work on Locality-aware Vision-Transformer is accepted at BMVC'23!

  - [Our work on Locality-aware Vision-Transformer is accepted at BMVC'23!](https://arxiv.org/pdf/2309.01561.pdf)

## Research

### AI Driven Digital Pathology for the 70 Gene Recurrence Signature: Development and Validation in Early Stage Breast Cancer

Agendia · Presented by Laura van ’t Veer

ESMO 2026 · LBA20 (Oral)

Our work at Agendia explores predicting recurrence risk directly from routine whole-slide images of breast cancer tissue, using supervision from the 70-gene signature.

- [Abstract](https://agendia.com/ai-driven-digital-pathology-for-the-70-gene-recurrence-signature-development-and-validation-in-early-stage-breast-cancer/)

- [Announcement](https://agendia.com/agendia-ai-driven-digital-pathology-esmo-2026/)

- [Publication on the website](https://kilickaya.github.io/#publication-esmo-2026)

### HyTAS: A Hyperspectral Image Transformer Architecture Search Benchmark and Analysis

Fangqin Zhou, Mert Kilickaya, Joaquin Vanschoren, Ran Piao

ECCV 2024

We present HyTAS, a quick method for hyperspectral vision transformer architecture search.

- [arXiv](https://arxiv.org/pdf/2407.16269)

- [GitHub](https://github.com/zhoufangqin/HyTAS/tree/main)

- [Publication on the website](https://kilickaya.github.io/#publication-1)

### Locality-Aware Hyperspectral Classification

Fangqin Zhou, Mert Kilickaya, Joaquin Vanschoren

BMVC 2023

We propose HyLITE, a vision-transformer for hyperspectral image classification.

- [arXiv](https://arxiv.org/pdf/2309.01561.pdf)

- [GitHub](https://github.com/zhoufangqin/HyLITE)

- [Publication on the website](https://kilickaya.github.io/#publication-2)

### Towards Label-Efficient Incremental Learning: A Survey

Mert Kilickaya, Joost van de Weijer, Yuki Asano

Preprint 2023

We survey semi-, few-shot, and self-supervised incremental learning.

- [arXiv](https://arxiv.org/pdf/2302.00353.pdf)

- [Slides](https://kilickaya.github.io/resume/survey.pdf)

- [GitHub](https://github.com/kilickaya/label-efficient-il)

- [Publication on the website](https://kilickaya.github.io/#publication-3)

### Are Labels Needed for Incremental Instance Learning?

Mert Kilickaya, Joaquin Vanschoren

CVPRW 2023 (Oral)

We introduce VINIL, a self-supervised incremental instance learner.

- [arXiv](https://arxiv.org/pdf/2301.11417.pdf)

- [Slides](https://kilickaya.github.io/resume/slides_vinil.pdf)

- [Publication on the website](https://kilickaya.github.io/#publication-4)

### Human-Object Interaction Detection via Weak Supervision

Mert Kilickaya, Arnold Smeulders

BMVC 2021

We introduce Align-Former to detect human-object interactions from an image.

- [arXiv](https://arxiv.org/pdf/2112.00492.pdf)

- [Slides](https://kilickaya.github.io/resume/slides_bmvc21.pdf)

- [Publication on the website](https://kilickaya.github.io/#publication-5)

### Structured Visual Search via Composition-aware Learning

Mert Kilickaya, Arnold Smeulders

WACV 2021

We introduce composition-aware learning to improve visual image search.

- [arXiv](https://openaccess.thecvf.com/content/WACV2021/papers/Kilickaya_Structured_Visual_Search_via_Composition-Aware_Learning_WACV_2021_paper.pdf)

- [Slides](https://kilickaya.github.io/resume/slides_wacv21.pdf)

- [Publication on the website](https://kilickaya.github.io/#publication-6)

### Re-evaluating Automatic Metrics for Image Captioning

Mert Kilickaya, Aykut Erdem, Nazli Ikizler-Cinbis, Erkut Erdem

EACL 2017

We propose Word-Mover Distance to evaluate image captioning.

- [arXiv](https://arxiv.org/pdf/1612.07600.pdf)

- [Slides](https://kilickaya.github.io/resume/slides_eacl17.pdf)

- [Publication on the website](https://kilickaya.github.io/#publication-7)

## Patents

### Visual Image Search via Conversational Interaction (Huawei)

Mert Kilickaya, Baiqiang XIA

[WO Patent, 2022](https://patents.google.com/patent/WO2023117041A1/)

### Network for Interacted Object Localization (Qualcomm)

Mert Kilickaya, Arnold Smeulders

[US Patent, 2022](https://patents.google.com/patent/US20220414371A1/)

### Subject-Object Interaction Recognition Model (Qualcomm)

Mert Kilickaya, Stratis Gavves, Arnold Smeulders

[US Patent, 2022](https://patents.google.com/patent/US11481576B2/)

### Context-driven Learning of Human-object Interactions (Qualcomm)

Mert Kilickaya, Noureldien Hussein, Stratis Gavves, Arnold Smeulders

[US Patent, 2021](https://patents.google.com/patent/EP4058930A1/)

## People & mentorship

- [Evangelos Tsimpouris](https://www.linkedin.com/in/evangelos-tsimpouris-141a75190/?originalSubdomain=nl) — MSc, TU/e, 2023-2024. [Learning to Transfer for Continual Learning](https://www.linkedin.com/in/evangelos-tsimpouris-141a75190/?originalSubdomain=nl)

- [Fangqin Zhou](https://research.tue.nl/en/persons/fangqin-zhou) — PhD, TU/e, 2022-2024. [Locality-Aware Vision Transformers](https://arxiv.org/abs/2309.01561v1)

- [Ran Piao](https://www.linkedin.com/in/ran-piao-a0a1a6281/?originalSubdomain=nl) — MSc, TU/e, 2022-2023. [Rapid Transformer Architecture Search without Training](https://drive.google.com/file/d/10ntRevPI_mF00oRE9_LhQrNr09hnHNuC/view)

- [Tommie Kerssies](https://scholar.google.com/citations?user=rrJn-eEAAAAJ&hl=en) — MSc, TU/e, 2022-2023. [Test Time Adaptation](https://arxiv.org/pdf/2208.08767.pdf)

- [Ceren Gok-Yildirim](https://research.tue.nl/en/persons/ceren-g%C3%B6k) — PhD, TU/e, 2022-2023. [Learning to Adapt for Continual Learning](https://arxiv.org/pdf/2303.13113.pdf)

- [Kishan Parshotam](https://scholar.google.com/citations?user=75hA5MQAAAAJ&hl=en) — MSc, UvA, 2020-2021. [Continual Learning of Object Instances](https://openaccess.thecvf.com/content_CVPRW_2020/papers/w15/Parshotam_Continual_Learning_of_Object_Instances_CVPRW_2020_paper.pdf)

- [Tarun Krishna](https://krishnatarun.github.io/) — MSc, UvA, 2019-2020. [Learning to rotate, scale and shift visual objects via disentanglement](https://drive.google.com/file/d/1klDGsZJFHzIJ1v2yQc-pCVPJESA-Rl8p/view)

## Beyond research

[Personal photography and visual memories](https://kilickaya.github.io/images.html)

## Source notes

- Research goals describe a direction of work, not a claim of solved or deployed capabilities.

- Dates retain their available precision. A null endDate for a current role means ongoing; other null fields mean no value is supplied on the website.

- The CV is an existing PDF and may lag the career information on the homepage.

- The ESMO 2026 abstract is linked to Agendia’s publication page. Its presenter is credited separately; the full author list and journal/DOI record are not supplied.

- The publication archive is the archive on this website, not a claim to include every subsequent work.

- Patent citation labels are preserved from the original archive and are not independently verified jurisdiction or grant dates.