| CARVIEW |
- Ph.D. candidate at the EPFL NLP Lab. Supervised by Prof. Antoine Bosselut.
- Goal: Building reasoning agents with a broad range of human cognitive abilities (learn, reason, plan, problem-solving, discovery) to serve the common good.
- Research: Large-scale AI Development (foundation & frontier), Learning & Reasoning Algorithms, World Model, and Agentic AI for Research & Science.
- Priors: Research Scientist Intern at Meta FAIR Lab (2024, 2025); Research Scientist Intern at Allen Institute for AI (AI2, 2022)
- Contact: Please feel free to say hi through my email, twitter, or LinkedIn!
About me
Hello! I am a PhD student at EPFL NLP Lab, supervised by Dr. Antoine Bosselut. My boradly study Machine Learning, NLP, and Multimodality.
My research aims to build AI reasoning agents that serve the common good with a focus on developing strong foundation models, designing novel reasoning paradigms, and adapting them to human-centered applications in critical areas like medicine and global fairness. Topics that I focus on include: Large-scale AI Development (Foundation & Frontier models), Learning & Reasoning Algorithms, World Model, and Agentic AI for Research & Science. My current work involves reframing AI reasoning as test-time continual learning with new methodologies in RL post-training and Deep Research agents.
I was a research scientist intern at Meta FAIR lab , supervised by Barlas Oğuz in 2025 and Ramakanth Pasunuru and Asli Celikyilmaz in 2024. I was also research scientist intern at Allen Institute for AI (AI2), supervised by Kyle Richardson and Ashish Sabharwal.
I received my bachelor's from Rose-Hulman Institute of Technology, with a double major in Computer Science and Mathematics. I started my research journey with Larry Moss (Indiana University) and Michael Wollowski.
- Learning & Reasoning Algorithms: A crucial challenge for the current AI systems is "How to achieve continual learning?" I research on new methods for test-time learning, a paradigm where models continue to learn from new information and experiences at inference time. I am interested in reformulating AI reasoning as test-time learning with applications to RL post-training and Deep Research. Under this paradigm, I am also interested in developing latent-space reasoning algorithms that operates over latent states optimizable at test time.
- Large-scale AI Development (Foundation & Frontier models): I am interested in developing next-generation large language, multimodal, & reasoning models, and their applications to various professional domains and tasks.
- World Model: Building spatially intelligent AI is the next decay of research advances. World model is a new type of generative models with the capabilities of understanding, reasoning, simulating and interaction with semantically, physically, geometrically and dynamically complex worlds. My research focuses on developing world models through large-scale video learning and simulation.
- Agentic AI for Research & Science: Deep Research agents are advanced AI systems driven by powerful foundation models and reasoning algorithms that can solve complex and novel problems. Their cabilities include the generation of novel ideas, effective information gathering through search tools, and the execution of analyses or experiments prior to drafting research reports or papers. My research aims to build research agents that can discover new knowledge through test-time continual learning.
In my free time, I enjoy doing many different things! Work-life balance is important for everyone's scientific career.
I am a musician. I play five different instruments: piano, flute, saxophone, piccolo, and cello. I enjoy playing classic music.
I also love to play tennis and lift weights in the gym.
I love traveling, visiting different places, and experiencing different cultures.
Distilling Counterfactual Data from Large Language Models
