Diagnosing Component-Level Failures in Computer-Use Agents
COLM 2026 (In review)Tianchen Guan, Daisy Xinlei Lin, Xiangjun Wang, Shuyan Zhou
Daisy Xinlei Lin
A scientist who builds. An engineer who designs.
Research Scientist & Engineer — Amazon AGI SF Lab
I build products that make people genuinely more capable with AI — not just more engaged. At Amazon AGI SF Lab, I owned the end-to-end RL training recipe for Nova Act, Amazon's production browser-use agent.
My PhD at NYU studied how humans plan and develop skill at complex tasks — and what makes some people get dramatically better while others plateau. I design AI interactions grounded in that research.
Right now I'm exploring new interaction paradigms for human-AI collaboration: tools that adapt to how you think, not the other way around.
What I've been building
The day job
Amazon Nova Act
ProductionOwned the end-to-end RL training recipe for Nova Act. Designed and shipped an online data curator that improved RL sample efficiency. Diagnosed and resolved a training/inference mismatch that stabilized RL training at scale across agentic tasks. Drove measurable gains in training data diversity and reliability.
Reinforcement Learning Computer Use LLM Post-Training
Read AWS Blog Post
The papers
Limitations in Planning Ability in AlphaZero
NeurIPS 2024Daisy Xinlei Lin, Brenden Lake, Wei Ji Ma
PaperLearning How Humans Learn to Play Board Games with GPT-4
AAAI Workshop 2024Veronica Yeom-Song, Xinlei Lin, Ionatan Kuperwajs, Heiko Schütt, Wei Ji Ma, Luigi Acerbi
Comparing Machine and Human Learning in a Planning Task
CogSci 2022Zheyang Sam Zheng*, Xinlei Daisy Lin*, Jake Topping*, Wei Ji Ma
PaperHuman Planning in Stochastic Environments
PsyArXiv 2025Jordan Lei, Jeroen Olieslagers, Nastaran Arfaei, Xinlei Lin, Wei Ji Ma
PaperComplementary cortical and thalamic contributions to cell type–specific striatal activity dynamics during movement
Science Advances 2026Enida Gjoni, Ram Dyuthi Sristi, Haixin Liu, Shahar Dror, Xinlei Lin, Keelin O'Neil, Oscar M. Arroyo, Sun Woo Hong, Hannah Kim, Jeffrey Liu, Sonja Blumenstock, Byungkook Lim, Gal Mishne, Takaki Komiyama
PaperThings that seem true to me right now ✱
- Women are controlled by hormones more than anyone talks about. Every week I'm a different person. It's the most underestimated battle.
- After walking 46 miles in Patagonia — got sick on day 3, had to recover overnight and keep going — I realized the human body's potential is basically unlimited.
- You can run a marathon untrained. I ran the NYC Marathon in 2024 with barely any training, finished in about 6 hours. You just have to trust yourself and vibe with the incredibly encouraging New Yorkers.
- Everything in moderation is the best philosophy.
- My entire PhD could be done in two months with the tools we have today. That's not a complaint — it's the most exciting thing about this moment.
- Making something you can hold — ceramics, tiramisu — after months of only making things in code is satisfying in a way I can't fully explain.
- Teaching my grandparents to use AI is the hardest product problem I've ever worked on.
- The best skincare routine is sunscreen and sleep.
- Tiramisu is the perfect dessert. I will not be taking questions.
When I'm not training agents
I snowboard when I can — Hokkaido is the dream — bake when I need to think, make ceramics, and am currently being managed by two cats named Boba and Milk who take their supervisory roles extremely seriously.
People ask about my skincare routine more than my research, which I find extremely funny. Happy to talk about both.