Unsafe2Safe: Controllable Image Anonymization for Downstream Utility
About me
Hello, my name is Minh. I’m a graduate student in Computer Science at Dartmouth College 🇺🇸, where I focus on generalizable AI systems, multimodal learning, and data-efficient representation within JinLab under the advising of Professor SouYoung Jin. My current work encompasses deepfake detection, instruction tuning, and vision-language modeling, with a particular interest in building methods that remain robust and useful across unseen domains.
I earned a Bachelor's degree with honors in Data Science from Aalto University 🇫🇮. I spent much of that time in Professor Stephane Deny's BRAIN Lab, where I worked as a research assistant on self-supervised learning and invariant representation. Alongside that, I completed my bachelor's thesis in scene representation with Professor Alexander Ilin and TA-ed advanced courses, including lectures hosted with Professor Nitin Sawhney.
In my final year, I went on exchange to the IC Department at EPFL 🇨ðŸ‡. The exchange was selective, fully funded, and combined course work with research. There, I explored diffusion-based deepfake detection and submitted a method to the ELSA Deepfake Detection Challenge that evaluates residual inversion signals.
I originally come from Hanoi, Vietnam 🇻🇳, where I studied at HNUE High School for Gifted Students and developed my passion for photography 📸.
Recent highlights
- Unsafe2Safe was accepted to the MUV Workshop on machine unlearning at CVPR 2026.
- The webpage and code for Unsafe2Safe are now available: project page.
- Unsafe2Safe was accepted to CVPR 2026.
- My paper on Latent Equivariant Operators for Robust Object Recognition: Promises and Challenges was accepted to an ICLR workshop. arXiv:2602.18406.
- Unsafe2Safe was presented as a poster at NECV 2025.
- Impact of Noise on Deep Learning-Based Pseudo-Online Gesture Recognition with High-Density EMG appeared at EMBC 2025.
- I successfully defended my master's thesis.
Research Exprience
I’m broadly interested in making AI systems more efficient, generalizable, and responsible.
My research has moved across a few connected directions, but the central question has stayed consistent: how can we build machine learning systems that remain useful, robust, and responsible outside the narrow setting they were trained in? At JinLab at Dartmouth, I work on privacy-preserving learning for video and multimodal models. Earlier, at Aalto University’s BRAIN Lab, I focused on invariant self-supervised representation learning, and during my research exchange at EPFL IVRL, I explored visual forensics and synthetic image detection.
Work Experience
Machine Learning Engineer · Pokedata
I also enjoy bringing machine learning systems into real products. At Pokedata, I have worked as a Machine Learning Engineer to build and productionize a fine-grained visual search system for mobile use.
The work spans detection, embedding, OCR, retrieval, and on-device optimization, with a strong focus on making the pipeline fast and robust enough for real-world use on iOS and Android.
Teaching Exprience
Teaching has been one of the most meaningful parts of my academic work. At Dartmouth and Aalto University, I have supported courses in machine learning, artificial intelligence, statistical inference, databases, and programming, often in roles that combined tutorials, mentoring, project guidance, and lecture support.
Fall 2025
COSC 274Grad-OnlyStatistical Inference and Machine Learning(Prof. Yaoqing Yang)
Fall 2025
COSC 70Foundations of Applied Computer Science(Prof. Adithya Pediredla)
Spring 2025
COSC 78/278Deep Learning(Prof. Yujun Yan)
Spring 2025
COSC 89.35/189Human-centered LLMsLead TA(Prof. Sarah Preum)
Spring 2025
COSC 74/274Statistical Inference and Machine Learning(Prof. Soroush Vosoughi)
Winter 2024
COSC 74/274Statistical Inference and Machine Learning(Prof. Soroush Vosoughi)
Winter 2024
COSC 70Foundations of Applied Computer Science(Prof. Yaoqing Yang)
Fall 2024
COSC 76/276Artificial Intelligence(Prof. Soroush Vosoughi)
Fall 2024
ENGS 108Applied Machine LearningLead TA(Prof. George Cybenko)
Spring 2023
CS-A1155Databases for Data ScienceLead TA(Prof. Nitin Sawhney)
Spring 2023
CS-E4800Artificial Intelligence(Prof. Jussi Rintanen)
Autumn 2022
ELEC-A7151Object oriented programming with C++(Prof. Yusuf Ali)
Autumn 2022
CS-C3240Machine Learning D(Prof. Pekka Marttinen & Prof. Stephan Sigg)
Service
International Journal of Computer Vision (IJCV)
I have served as a reviewer for the journal.
2022 IEEE International Symposium on Information Theory (ISIT)
I volunteered at ISIT 2022, where I supported all plenary sessions and more than five tutorial sessions, helping prepare the stage and assist speakers with their presentations.