My (Chiffon) Nguyen she/her

Nguyễn Trà My / 阮沐茶 / 윈자미

AI Research for Broader World & Life-long Learning

Pronunciation

My = /me/, Chiffon = /shi-FON/. I'm Chiffon in English-speaking context.

I research current and future AI for broader world:

Pluralistic alignment, and more broadly AI Safety & trustworthy AI
Multilingual & multicultural AI
Uplifting human agency for future of education and work with AI
AI impacts on animals and ecology, including animal welfare, climate change
AI for Science

For technical research, I’m interested in AI measurement science↗ and mechanistic understanding to make grounded, predictive, specific claims about AI capability and safety.

I’m interning at Lida Safety Research↗, contributing to projects in persona alignment, scalable oversight, and sandbagging. I also co-lead Cohere Labs Open Science Community’s multi-cultural riddles analysis subteam. Recently I worked on SEATauBench↗ and CoT monitoring.

I am applying to research-oriented Master’s in Computer Science, AI, NLP, or related fields for Spring 2027 and seeking research opportunities (internships, research assistantships, collaborations). For research collaborations, drop me an email!

Career highlights

see CV →

Feb 2026 - Present
AI Research Mentee (Multilingual Agentic Evaluation) @ SEACrowd · Remote

Evaluating conversational agents on Tau2-Bench across five settings with increasing language and cultural adaptation. Mentors: Dr. Samuel Cahyawijaya (Cohere) & Patomporn Payoungkhamdee (VISTEC). In Submission
Oct 2025 - Present
AI Research Mentee (Chain-of-thought Monitoring) @ Algoverse AI Research · Remote

Lead project studying the boundaries of weak-to-strong chain-of-thought monitoring on safety-relevant evals (WMDP & SWEBench-Verified) using sandbagging as a case study.
Sep 2021 - May 2025
B.Sc in Computational Sciences @ Minerva University · San Francisco, CA, USA (+ Seoul, Taipei, Hyderabad, Buenos Aires, Berlin)

Relevant Coursework: Machine Learning, Bayesian Modeling, Statistical Modeling & Causal Inference, Optimization Methods.
Jun 2024 - Aug 2024
ML Research Assistant @ Landshut University of Applied Sciences · Landshut, Bavaria, Germany

3D object detection in point clouds and point cloud registration, supervised by Prof. Eduard Kromer and Prof Sandra Eisenreich

Latest updates

see all →

Jun 4, 2026

Started interning at Lida Safety Research↗, contributing to projects in persona alignment, scalable oversight, and sandbagging.

May 1, 2026

Started co-leading [Cohere Labs Open Science Community]‘s multi-cultural riddles benchmarking project with other six leads, working with 100+ community researchers.

Apr 15, 2026

Started ElectricSheep’s Futurekind AI Fellowship Spring 2026↗ where I research and build solutions in the intersection of AI and animal protection.

Selected publications

see all →

SEATauBench: Adapting Tool-Agent-User Evaluation Into Low-Resource Southeast Asian Languages

My Chiffon Nguyen^*, Aulia Adila^*, Saksorn Ruangtanusak^*, Kittiphat Leesombatwathana^*, Vissuta Gunawan Lim^*, Patomporn Payoungkhamdee, Samuel Cahyawijaya,

In Submission, Pending Preprint • 2026

Abstract

We introduce SEATauBench, the first agentic-focused evaluation framework for sovereign AI development in Southeast Asia, a region of strategic importance with over 700 million people. Despite growing regional evaluation efforts, existing multilingual agents show limited capability when operating in SEA languages, particularly in mixed-language scenarios. Through evaluation across multiple adaptation approaches, we find that while English agentic capabilities transfer to target language responses, performance degrades significantly when context is provided in SEA languages. We propose a translation-based mitigation strategy that preserves entity consistency while enabling agents to leverage English comprehension. SEATauBench establishes a rigorous benchmark for sovereign AI agent assessment, providing diagnostic tools to address capability gaps and support agentic AI development in diverse linguistic communities in the region.

Miscellaneous

My Vietnamese name is ‘Trà My’, which means Camellia japonica↗ (tea flower).
I graduated in May 2025 from Minerva University↗ studying Machine Learning & Statistics, with peers from 50+ countries and living in 6 cities as part of its global immersion program↗: Seoul (South Korea), Chinese Taipei, Hyderabad (India), Buenos Aires (Argentina), Berlin (Germany). This experience heavily shapes my worldviews and motivates my research focus in diversity and collaboration.
Things that restore my energy: walks, reading↗, cooking, cycling, card reading, and astrology. I also enjoy solo travel, food, cultural activities, and historical museums.
Things that excite me outside research: teaching & mentoring, event planning and organizing, languages and cultures, philosophical and political discussions.
I’m learning Mandarin Chinese (band 4/B1) and French (A2).
I’m agnostic, with wonderful friends that are Christian, Buddhist, and Muslim.
I’m on no-meat diet at least 4 days/week, and trying to transition to vegetarian diet.
I like Rust-based and well-designed stuff. Recommend: Astral’s Python tooling↗ (uv, ruff, ty), Inspect↗ for reproducible LLM eval, LMMs-eval↗ for multimodal eval, Rust-based CLI (ripgrep↗, eza↗, bat↗). My hardware and apps here.