About Me

Machine Learning Engineer & Researcher
Efficient LLM Systems · Generative Models

I am Jeng-Yue (Buffett) Liu 劉正悦, a Master’s student in Artificial Intelligence and Innovation at Carnegie Mellon University. I work on efficient machine learning systems and generative models: I care about models that are not only capable, but also fast, controllable, and reliable enough to run in the real world.
Most of my recent work is on diffusion language models (dLLMs), which generate text by filling in many tokens in parallel rather than strictly left to right. I study how to decode them under constraints, how to check whether what they produce is actually correct, and how to serve them efficiently once the usual autoregressive assumptions no longer hold. Along the way I built dLLM Notes, a learning website for dLLMs that walks from the training objective all the way to production serving, with interactive demos and annotated paper notes.
I came to this from audio. At the Music and Audio Computing Lab, Academia Sinica, working with Prof. Yi-Hsuan Yang and Prof. Li Su, I developed SynthCloner (ICASSP 2026) for controllable synthesizer sound conversion, and later brought similar ideas into production at Neutone. Before CMU, I earned dual Bachelor’s degrees in Information Management and Geography from National Taiwan University, graduating summa cum laude.
My interests center on:

• Efficient LLM inference & serving systems
• Diffusion language models: decoding, evaluation, and serving
• Controllable music and audio generation

I am actively seeking full-time Software Engineering / Machine Learning Engineering roles starting May 2028, and I am always glad to connect with engineers and researchers working on similar problems.

Education

University LogoCarnegie Mellon University (CMU)
School of Computer Science
M.Sc. in Artificial Intelligence and Innovation
Aug. 2025 - May. 2028
  • Overall GPA: 3.88 / 4.0
  • Current Coursework: 26' Spring: Diffusion and Flow Matching (10-799), Advanced NLP (11-711), LLM Systems (11-868), Intro to Deep Learning (11-785), AI Engineering (11-695), AI Venture Studio (11-681)25' Fall: Intro to Machine Learning (10-601), Coding Bootcamp (11-601), Gen AI for Music & Audio (15-798), AI & Future Markets (11-651), Law of Computer Technology (17-762)25' Summer: Intro to Computer Systems (15-503)
University LogoNational Taiwan University (NTU)
Dual Degree:
B.B.A. in Information Management & B.S. in Geography
Sep. 2020 - Jun. 2025
  • Overall GPA: 3.95 / 4.3
  • CS-related GPA: 4.15 / 4.3
  • Honors: Summa Cum Laude — top 1% of the school

Awards

2025

Summa Cum Laude, National Taiwan UniversityGraduated with the highest Latin honors, awarded to the top 1% of the school.

2024

The Phi Tau Phi Scholastic Honor Society of the Republic of ChinaInducted as an honorary member, recognizing the top 1% of students for outstanding academic performance at the university.

2024

Bachelor Degree Thesis Award
Ranked top 3 in department

2023

NTU Presidential Award
Ranked 1/49 in department

2023

NTU Dean’s List Award
Twice