[Update 3/9/26] On my 25th birthday, I joined Nextdoor as a Machine Learning Engineer for my first full-time job, working on feed ranking. I’m now based in New York City (somewhere between Williamsburg and Bushwick, if we’re being specific) and still don’t know that many people here. Hit me up if you’re around!

-

Hi, I’m Hao-yung Weng, a Master’s student in the Machine Learning Department at Carnegie Mellon University. I also work in the Auton Lab at the Robotics Institute, advised by Prof. Artur Dubrawski. My research focuses on natural language processing, speech processing, and large language models (LLMs). Lately, I’ve been exploring LLM personalization and federated learning to improve adaptability while preserving privacy—a key challenge for scalable AI.

Beyond academia, I have diverse industry experience. This summer, I was an AI Research Intern at HeyGen, where I helped launch the flagship product, “Video Agent”, as part of the founding team. In short, it’s an agentic AI system that transforms prompts and media into dynamic, multi-scene videos. Among other contributions, I was in charge of the referencing module, enabling the agent to infer and reuse stylistic patterns—like tone, pacing, and shot structure—from reference videos, ideal for creators seeking consistency with minimal effort.

Previously, at Google, I developed a bug triage system that streamlined vendor collaboration under the Joint Development Manufacturer (JDM) model, cutting issue resolution time by 25%. At WorldQuant, I built mathematical models to predict equity market movements and designed novel, profitable trading strategies.

Before Carnegie Mellon, I graduated as valedictorian and summa cum laude (top 1%) in computer science from National Taiwan University. There, I worked with Professor Yun-Nung Chen on natural language processing and transfer learning, and with Professor Hung-yi Lee on speech processing and parameter-efficient fine-tuning.

In 2023, I was honored as one of Taiwan’s Outstanding Youth by the Ministry of Education.

When I’m not wrangling models, I’m a big coffee geek and write blog posts exploring the technical, scientific, and occasionally nerdy sides of brewing—like this one, where I evaluate Wet WDT, a technique that’s essentially stirring coffee grounds with—well—acupuncture needles.

Cheers, and take care! Maybe give my favorite post-rock band a listen?