What’s up! I’m Jason, a 3rd-year undergrad studying computer engineering at UC San Diego. I am currently working under the supervision of Professor Hao Zhang @ Hao AI Lab and Professor Tajana Šimunić Rosing @ SeeLab. I am currently involved with a several cutting-edge LLM research, more specifically on LLM evaluation, compression, and KV-caches.

Why Research?

My Values and Attributes: Foresight and initiative are two key attributes that I practice every waking second. I channel my foresight into choosing the right problems to solve, and I fuel my initiative by understanding the importance of my work.

My Community: A big part of what makes research so enjoyable for me is belonging to a wonderful research community and having friendships with my smart PhD colleagues. If you’d like to chat, feel free to connect with me on 𝕏.

Projects and Publications

TinyAgentQuantization-aware Model Compression and Adaptation for On-device LLM Agent Deployment

• We introduce a novel LLM edge deployment solution to automatically fine-tune and compress domain-specific LLMs with up to 8x memory saving and inference speedup with minimal performance loss. This work was presented at the ICML 2024 ES-FoMo-II workshop.

Awards

Best Poster Award - PRISM Annual Review

• I was awarded the best poster award for PRISM Annual Review for Theme 1 - Systems & Software. November 8, 2024.