Email
GitHub
Google Scholar
LinkedIn
Curriculum vitae
Geometry-Aware Trajectory State Modeling for Few-Shot Action Recognition
Geometry-aware trajectory modeling for recognizing actions from a small number of examples.
Marko Haralović
I’m a master’s thesis student in the Photogrammetry and Remote Sensing group at ETH Zürich, supervised by Li Mi and mentored by Konrad Schindler. We collaborate with the OATML group at Oxford. My thesis investigates geospatial knowledge and hallucinations in vision-language models.
I’m pursuing an MSc in Data Science at the University of Zagreb. Previously, I spent an exchange year and worked as a research assistant at the University of Twente with Estefanía Talavera, studying action understanding and surgical video analysis.
Publications & preprints
When Adaptation Hurts: Connecting Representational Drift to OOD Failures in MedSAM Fine-Tuning
Studying how adaptation changes medical foundation model representations and robustness under domain shift and imperfect prompts.
Beyond Instrument Motion: Recognizing Tissue Tension Toward Surgical Skill Assessment
Recognizing tissue tension in robot-assisted surgery with sparse point tracking, toward richer surgical skill assessment.
G3Ego: Gaze-Guided Graphs for Egocentric Action Understanding
Using gaze to build compact scene graphs focused on the hand-object interactions that matter for first-person action understanding.
CrisisKD: Five-Stage Knowledge Distillation for Aspect-Level Sentiment and Emotion Analysis in Crisis Discourse
Distilling teacher-generated reasoning into smaller language models for aspect-level sentiment and emotion analysis in crisis-related social media.
* Equal contribution
Research & industry
Sep 2026 — Apr 2027
Master’s Thesis Internship
ETH Zürich · Photogrammetry and Remote Sensing
2023 — Present
Founder
Computer Vision and Deep Learning Consultancy
Nov 2025 — Aug 2026
Research Assistant
University of Twente · EEMCS
Dec 2024 — Sep 2025
Multimodal Deep Learning Researcher
Arkensight
Mar 2025 — Jul 2025
Research Assistant
University of Zagreb · FER · Project for Google
Aug 2024 — Oct 2024
Computer Vision Research Intern
Visage Technologies
Jul 2024 — Aug 2024
Data Science Intern
mStart · Fortenova Group
Nov 2022 — Jul 2024
Junior Deep Learning Engineer, R&D
Ericsson
Jul 2022 — Sep 2022
Data Science Intern, R&D Summer Program
Ericsson
Education
Sep 2026 — Jul 2027 (expected)
ETH Zürich
Master’s thesis · Photogrammetry and Remote Sensing
Supervisor: Li Mi · Mentor: Konrad Schindler · Collaboration with the OATML group at Oxford
Sep 2024 — Jun 2027 (expected)
University of Zagreb
MSc in Data Science · FER
GPA 4.53 / 5.00 · Mentors: Siniša Šegvić, Dario Bojanjac
Sep 2025 — Jul 2026
University of Twente
MSc Data Science · Exchange year
GPA 8.50 / 10.00 · Estefanía Talavera
Sep 2021 — Jun 2024
University of Zagreb
BSc in Computer Science · FER
Mentor: Siniša Šegvić · Thesis: semantic segmentation of satellite images
Honors & awards
- 2026
FER Scholarship for International Research Mobility
One of five awarded · Thesis research at ETH Zürich
- 20261st place
Best Performance Award · ACDC Challenge
1st of 12 teams · Medical Image Analysis, University of Twente
- 2025
AMAC FER Alumni Scholarship for Academic Excellence
2025/2026 · Top 1%
- 2025
Rector’s Award · University of Zagreb
Methods for Assessing and Improving the Fairness of Deep Classifiers
- 20251st place
LUMEN Data Science Competition
1st of 30 teams · Fairness-aware melanoma classification
More awards & scholarships
- 3×
Conference Travel Grants
FER
- 20252nd place
HACKL Hackathon · City of Zagreb
2nd of 20 teams
- 2025
Erasmus+ Scholarship
Academic year 2025/2026
- 2024
Algotrade Hackathon
4th of 75 teams
- 2022
Ericsson Award for Best Student Work
2nd place · Academic year 2021/2022
- 2021
STEM Scholarship
Top 1% of high school students
- 2021
Croatian National Philosophy Competition
3rd of 30 finalists