Hüseyin Umut Işık

About

I am currently a research intern at the Quantitative Intelligent Medical Imaging (QUIN) research group at Boston Children's Hospital / Harvard Medical School, where I work on medical image analysis. My research interests are in computer vision and deep learning, in particular monocular depth estimation for lunar robotics and medical imaging.

On the side, quantum computing is a hobby I take seriously: my teams have won first place at the EPFL Quantum Hackathon (Quandela Challenge) and IQM QuantumHACK Madrid, and received an honorable mention at MIT iQuHACK.

News

• 2026 ARC-CT, our region-aware vision-language model for 3D chest CT, was accepted to the MICCAI 2026 TIA workshop.
• 2026 Selected for the venture track of Hack-Nation Venture Lab (Cohort 3).
• 2026 3rd place at the Hack-Nation Global AI Hackathon (Genome Firewall Challenge), among 2,000+ participants.
• 2026 Selected for Y Combinator Startup School 2026 in San Francisco.
• 2026 LuMon, our benchmark for lunar monocular depth estimation, was accepted to the CVPR 2026 AI4SPACE workshop.
• 2026 Started as a research intern at the QUIN research group, Boston Children's Hospital / Harvard Medical School.
• 2026 1st place at the EPFL Quantum Hackathon, Quandela Challenge.
• 2026 Honorable mention at MIT iQuHACK (NVIDIA Challenge), among 1,400+ participants.

Publications

• ARC-CT: Anatomy-Routed Contrastive Vision-Language Learning for 3D Chest CT 2026
  MICCAI 2026, TIA Workshop

  Hüseyin Umut Işık, Mehmet Alp Özaydın, Sıla Kurugöl, Şeyda Ertekin.

  A region-aware vision-language model for 3D chest CT: anatomy-routed contrastive learning aligns volumetric scans with radiology reports for region-level reasoning.

• LuMon: A Comprehensive Benchmark and Development Suite with Novel Datasets for Lunar Monocular Depth Estimation 2026
  CVPR 2026, AI4SPACE Workshop

  Aytaç Sekmen, Fatih Emre Güneş, Furkan Horoz, Hüseyin Umut Işık, Mehmet Alp Özaydın, Onur Altay Topaloğlu, Şahin Umutcan Üstündaş, Yurdasen Alp Yeni, Halil Ersin Söken, Erol Şahin, Ramazan Gökberk Cinbiş, Sinan Kalkan. arXiv

Experience

• Research Intern, QUIN Research Group 2026 – present
  Harvard Medical School / Boston Children's Hospital

  Medical image analysis research.

• Computer Vision Candidate Engineer 2025 – 2026
  MKE – Makine ve Kimya Endüstrisi

  Object detection, tracking, and mapping for unmanned ground vehicles (UGVs).

• Undergraduate Researcher 2024 – 2026
  METU ROMER

  Event-based vision and monocular depth estimation; simulation environments for robotic perception, including lunar applications.

• Software Team Member 2024 – present
  METU ANZU UAV

  Autonomous UAV systems: path planning, real-time object detection, and tracking.

• R&D Intern 2025
  HIDROAN

  AI model optimization: RF-DETR object detection deployed with TensorRT, ONNX, and PyTorch on embedded edge hardware.

Programs

• Y Combinator Startup School 2026 Jul 2026
  San Francisco, CA

  Selected from 30,000+ applicants for a hand-picked group of founders and technical builders; two days of sessions with Jensen Huang, Sam Altman, Alexandr Wang and more.

• Hack-Nation Venture Lab, Cohort 3 2026
  Remote

  Selected for the venture track on the strength of a 3rd place finish at the Hack-Nation Global AI Hackathon, an MIT-born founder program.

Selected Awards & Projects

• Hack-Nation Global AI Hackathon, 3rd place, Genome Firewall Challenge 2026

  Gatcha: a model predicting target antibiotics directly from bacterial genome sequences, built in team Qedi among 2,000+ participants. code · project

• CENG488 Guided Research Symposium, 1st place 2026

  Ranked 1st among 29 projects for Region-Aware CLIP for Chest CT, the work that became ARC-CT.

• EPFL Quantum Hackathon, 1st place, Quandela Challenge 2026

  Hybrid Photonic Temporal QRC: mapped non-linear market dynamics into a photonic quantum reservoir (MerLin) to forecast swaption volatility surfaces. It beat LSTM baselines with fewer parameters. Team Qedi: Eren Aslan, Arda Kara, Mehmet Alp Özaydın, H. Umut Işık. code · project

• MIT iQuHACK, Honorable mention, NVIDIA Challenge 2026

  Hybrid GQE-MTS with transfer learning for the Low Autocorrelation Binary Sequences (LABS) problem. The approach combines generative models with HPC. Team QAT: Hatice Boyar, Eren Aslan, İlayda Dilek, Jen-Yu Chang (Leo), H. Umut Işık. code · project

• IQM QuantumHACK, Madrid, 1st place, Banco Santander Challenge

  Variational quantum classifier for risky investments, run on IQM quantum hardware in a 24-hour hackathon.

• TEKNOFEST Quantum Technologies, 1st place

  Quantum neural networks for low-quality face recognition, 30-hour hackathon in Istanbul.

• TÜBİTAK 2204-A, 1st place, Izmir regional finals

  Cryptography algorithms and mathematics project.

• METU-DTX Hackathon, 2nd place

  Industrial defect classification model built from scratch in 6 hours.

• BTK Academy CTF, 2nd place

  4-day CTF camp on web application security and kernel vulnerabilities.

Miscellaneous

Outside of work I draw, play football, solve Rubik's cubes, and run. My current goal is to finish an Ironman.