Education
M.Eng (Research)
Singapore University of Technology and Design
Expected 2025 Sep to 2026 Sep
- Cumulative GPA: 5.00/5.00
- AI Singapore Accelerated Masters Scholar (2024 Sep - 2026 Sep)
- DSO-AISG Research Award (2025 Sep - 2026 Sep)
B.Eng (Computer Science, Minor in AI)
Singapore University of Technology and Design
2021 Sep to 2025 May
- Cumulative GPA: 4.70/5.00, Honors with Highest Distinction
- SUTD Undergraduate Merit Scholar (2021 Sep - 2024 Aug)
Global Exchange Program
Aalto University
Finland, Jan 2024 to Jun 2025
- Awarded the KKH Global Exchange Award (2024 Spring)
- Exchange Student under Aalto University’s School of Science
| Relevant Courses | Grade |
|---|---|
| 51.511 Multimodal Generative AI – Postgraduate | A |
| 99.512 Mathematics for AI – Postgraduate | A |
| 51.504 Machine Learning – Postgraduate | A |
| 50.050 Advanced Algorithms (Formerly Discrete Mathematics and Algorithm Design) | A |
| 50.035 Computer Vision | A |
| 50.007 Machine Learning - Undergraduate | A |
| 01.117 Brain-Inspired Computing and it’s Applications | A |
| CS-E4890 Deep Learning | Exchange |
| ELEC-E5550 Statistical Natural Language Processing | Exchange |
| CS-E4800 Artificial Intelligence | Exchange |
Work Experience
Research Intern, DSO-AISG Research Award
2025 September – Present
- Collaborated with research scientists from DSO National Laboratories as part of the research award to improve VLM robustness against ambiguous and corrupted modalities.
- Applied the Partial Information Decomposition (information theoretic) framework to motivate hypotheses and design experiments for tuning multimodal interactions.
- Curated training (e.g. cleaning, deduplication) and validation partitions using vLLM, yielding three training sets (984,000 samples) of varying redundant interactions for supervised fine-tuning.
- Tested the hypotheses by fine-tuning VLMs — ranging from 256M (SmolVLM) to 8B (LLaVa-OneVision) parameters — with low rank adapters, leading to a 38.3% decrease in visual-induced hallucinations and 16.8% gain in consistency; methods and findings were accepted into the ICML main track.
Founding AI Engineer, Pallo (formerly Check) (Iterative W25)
2025 January – March
- Built a Retrieval Augmented Generation (RAG) workflow to produce syllabus-accurate outputs, securing a Pre-Seed fund from Iterative VC (Winter 2025).
- Built a data processing pipeline by combining open-source Computer Vision models with LLMs, contributing 8,700 high quality questions to Supabase for RAG.
- Deployed DeepSeek R1 models to Google Cloud Run to solve Singapore GCE A-Level math problems, increasing the accuracy of the final outputs by 36%.
LMM Research Assistant, Social AI Lab (SUTD)
2024 June – December
- Web scraped using Selenium and Beautiful Soup to collect more than 28,000 comics for analysis.
- Recruited and managed 8 participants to evaluate 2,800 comics using Label Studio to assess large multimodal models’ (LMMs such as Qwen2-VL, LLaVa-OV, GPT4o and Gemini) ability to comprehend humor.
LLM Research Intern, DSO National Laboratories
2023 August – December
- Applied the Graph of Thoughts reasoning workflow with LLMs to detect vulnerable code within a 3-layer call stack, reducing incurred API (ChatGPT) costs by 25%.
- Integrated Llama 2 and Code Llama with LangChain to perform Retrieval-Augmented Generation (RAG), further improving contextual understanding.
Technical Skills: Python, PyTorch, TensorFlow, Google Cloud Platform, SQL, Java
Publications
Conference Papers
Self-Captioning Multimodal Interaction Tuning: Amplifying Exploitable Redundancies for Robust Vision Language Models
43rd Proceedings of the International Conference on Machine Learning, Seoul, South Korea (ICML 2026) · 2026
@inproceedings{ ryan2026,
title = { Self-Captioning Multimodal Interaction Tuning: Amplifying Exploitable Redundancies for Robust Vision Language Models },
author = { Yuriel Ryan and Ip Hei Man and Adriel Kuek and Paul Pu Liang and Roy Ka-Wei Lee },
booktitle = { 43rd Proceedings of the International Conference on Machine Learning, Seoul, South Korea (ICML 2026) },
year = { 2026 }
}Humor in Pixels: Benchmarking Large Multimodal Models Understanding of Online Comics
Findings of the Association for Computational Linguistics, Suzhou, China (EMNLP 2025) · 2025
@inproceedings{ ryan2025,
title = { Humor in Pixels: Benchmarking Large Multimodal Models Understanding of Online Comics },
author = { Yuriel Ryan and Rui Yang Tan and Kenny Tsu Wei Choo and Roy Ka-Wei Lee },
booktitle = { Findings of the Association for Computational Linguistics, Suzhou, China (EMNLP 2025) },
year = { 2025 }
}Workshop Papers
Balance Human Agency & AI Assistance in the Tussle for the ``Right’’ to Choose, Own, Work, and Learn
Trustworthy AI for Good (AI4GOOD) Workshop, Seoul, South Korea (ICML 2026) · 2026
@inproceedings{ khoo2026,
title = { Balance Human Agency & AI Assistance in the Tussle for the ``Right'' to Choose, Own, Work, and Learn },
author = { Zi-Yu Khoo and Yuriel Ryan and Nicole Heng Yim Oo and Hui En Pang and Eric J. W. Orlowski and Hakim Norhashim and Ruth Wan Theng Chew and Davin Choo and Rachael Hwee Ling Sim and Simon Chesterman and Jungpil Hahn and Bryan Kian Hsiang Low },
booktitle = { Trustworthy AI for Good (AI4GOOD) Workshop, Seoul, South Korea (ICML 2026) },
year = { 2026 }
}