About Me

Hello! I’m a third-year Ph.D. student at UC Berkeley, advised by Prof. Matei Zaharia and affiliated with the Sky Computing Lab. I also work closely with Prof. Ion Stoica and Prof. Joey Gonzalez. I design and build efficient and reliable machine learning systems through joint optimization across the systems stack. Broadly, I’m interested in:

  1. Responsible and efficient frameworks for emerging agent systems, with sustainability as a first-order objective
  2. Reliably harnessing LLMs to accelerate scientific discovery in systems research and beyond
  3. Machine learning systems and data center computing

Before coming to Berkeley, I completed my Master’s degree at Carnegie Mellon University (CMU), where I was very fortunate to work with Prof. Akshitha Sriraman. My Master’s research centered on developing sustainable resource management techniques for microservices in hyperscale data centers.

I graduated from the University of Toronto with a Bachelor’s degree in Computer Engineering. While there, I was advised by Prof. Steve Mann on wearables for prosopagnosia. In between my academic pursuits, I spent ~3 years as a software engineer at IBM working on Db2 database core engine development, specifically on availability features such as backup, restore, and recovery.

Please feel free reach out for research, collaborations, or a casual chat, especially if you are a junior, disadvantaged, or underrepresented student. The best way to reach me is via email: [melissapan]@berkeley.edu

CV: PDF here, up to Aug 2026.


News

2026 Gave 15 invited talks so far this year: Measuring Agents in Production at Microsoft Azure Research, NVIDIA, Amazon, UW SyFI Lab, AGI House, NICE, and more; MAST at LangChain (hosted by Harrison Chase), the University of Oxford & the Alan Turing Institute, and Edge Delta; BRANE at Stanford and Arena Intelligence; and Barbarians at the Gate at Meta.
Jul '26 Presented Measuring Agents in Production as an Oral & Spotlight paper at ICML 2026 in Seoul! Also featured in Nebius Science’s ICML 2026: Papers That Matter.
Jul '26 Fantastic Adaptive Taxonomies and How to Use Them won the Best Paper Award at the ICML 2026 Workshop on Failure Modes of Agentic AI!
Jul '26 I am interning at Arena Intelligence this summer as a machine learning research intern, working with Prof. Ion Stoica on coding agent efficiency.
May '26 New preprint: Natural Language Query to Configuration for Retrieval Agents (BRANE), a Laude Slingshots project, with our collaborators at Microsoft Azure Research.
May '26 SkyDiscover is presented at ACM CAIS 2026 as a Spotlight in the Demo Track!
Mar '26 My work on energy-efficient AI systems is featured in The Daily Californian: UC Berkeley researchers use AI to tackle climate problems.
Feb '26 MAP: Measuring Agents in Production is selected for the Laude Institute’s Slingshots // TWO batch!
Dec '25 I am selected as a Laude Institute AI Resident!
Dec '25 Presented MAST (Spotlight) at NeurIPS 2025 in San Diego, and Towards Automatically Optimizing Retrieval Augmented AI Systems at the ML for Systems workshop!
Dec '25 New preprint: Measuring Agents in Production — a study of how LLM agents are deployed and measured in production across industry.
Sept '25 MAST is accepted to NeurIPS 2025 as a Spotlight paper in the Dataset and Benchmark Track! Top 2.8% of all submissions and top 11.26% of all accepted papers!
Aug '25 I am honoured to receive the Amazon AI Fellowship.
July '25 Presented Electro at OSDI Poster Session. Link
Summer '25 Invited talks for MAST at MIT DSG, IBM Watson Lab, Berkeley Agentic AI Summit, Microsoft Asia Research, Accenture, Berkeley SkyDeck Demo Day, BAIR-RDI Agent Workshop, and 7 other organizations.
Oct '24 Presented at SOSP24 poster session, and SYSDW workshop!
Aug '24 Joined UC Berkeley EECS! Honoured to receive the Katherine & James Lau EECS Fellowship.
May '24 Worked on graph optimization in edge TPU at Google.
May '24 Graduated from CMU MSECE!
May '23 Worked on graph optimization in DTensor at Google.
Jan '23 I started my graduate study at CMU!

Selected Research Projects

Melissa Z. Pan*, Negar Arabzadeh*, Riccardo Cogo, Yuxuan Zhu, Alexander Xiong, Lakshya A. Agrawal, Huanzhi Mao, Emma Shen, Sid Pallerla, Liana Patel, Shu Liu, Tianneng Shi, Xiaoyuan Liu, Jared Quincy Davis, Emmanuele Lacavalla, Alessandro Basile, Shuyi Yang, Paul Castro, Daniel Kang, Koushik Sen, Dawn Song, Joseph E. Gonzalez, Ion Stoica, Matei Zaharia, Marquita Ellis (* equal contribution)
ICML 2026
Published Oral Spotlight
Mert Cemri*, Andrei Cojocaru*, Melissa Z. Pan, Shu Liu, Shubham Agarwal, Alexander Krentsel, Jay Tang, Kannan Ramchandran, Joseph E. Gonzalez, Matei Zaharia, Alex Dimakis, Ion Stoica (* equal contribution)
ICML Workshop on Failure Modes of Agentic AI 2026
Published Best Paper
Shu Liu, Mert Cemri, Shubham Agarwal, Alexander Krentsel, Ashwin Naren, Qiuyang Mang, Zhifei Li, Akshat Gupta, Monishwaran Maheswaran, Audrey Cheng, Melissa Z. Pan, Ethan Boneh, Kannan Ramchandran, Koushik Sen, Matei Zaharia, Alexandros G. Dimakis, Ion Stoica
ACM CAIS, Demo Track 2026
Published Spotlight
Melissa Z. Pan, Negar Arabzadeh, Mathew Jacob, Fiodar Kazhamiaka, Esha Choukse, Matei Zaharia
arXiv preprint 2026
Preprint
Why Do Multi-Agent LLM Systems Fail?
Melissa Pan*, Mert Cemri*, Shuyi Yang*, Lakshya A. Agrawal, Bhavya Chopra, Rishabh Tiwari, Kurt Keutzer, Aditya Parameswaran, Dan Klein, Kannan Ramchandran, Matei Zaharia, Joseph E. Gonzalez, Ion Stoica (* equal contribution)
NeurIPS Datasets and Benchmarks Track 2025
Published Spotlight
The first empirical, principled study of multi-agent LLM failures. We introduce MAD (1,242 annotated traces across 7 frameworks) and MAST, a taxonomy for common failure patterns that currently exist in multi-agent LLM systems.
Melissa Z. Pan, Negar Arabzadeh, Mathew Jacob, Fiodar Kazhamiaka, Esha Choukse, Matei Zaharia
Machine Learning for Systems Workshop @ NeurIPS; also presented at the OSDI 2025 Poster Session 2025
Published
Melissa Pan*, Audrey Cheng*, Shu Liu*, Zhifei Li, Bowen Wang, Alex Krentsel, Tian Xia, Mert Cemri, Jongseok Park, Shuo Yang, Jeff Chen, Lakshya Agrawal, Aditya Desai, Jiarong Xing, Koushik Sen, Matei Zaharia, Ion Stoica (* equal contribution)
arXiv preprint 2025
Preprint

Teaching & Mentorship & Services

I’m passionate about education and supporting the next generation of (computer) scientists, especially students from underrepresented backgrounds.

Teaching:

  • Expanding Your Horizons, Creative Coding Workshop (Spring 2025, Spring 2026) - Designed and led a 1-day workshop for girls in the area from grade 5 - 9.
  • Invited by Dr. Brad Bass as Guest lectures on sustainable computing and ML systems for University of Toronto’s course ENV360.
  • 18-847: Data Center Computing (Spring 2024) - Graduate Teaching Assistant at Carnegie Mellon University
  • 18-847: Modern Computer Systems (Fall 2023) - Graduate Teaching Assistant at Carnegie Mellon University
  • STEM4Girls Program - Volunteer mentor for middle school girls interested in STEM

Mentorship:

  • Berkeley EAAA (2025): PhD application mentoring for underrepresented students
  • Berkeley EAAA (2024): PhD application mentoring for underrepresented students
  • Regular undergraduate research coffee chats (2024)
  • CMU ECE Graduate Mentorship Committee Chair (2024)

Academic Services:

  • NeurIPS Reviewer (2026)
  • ACM CAIS Reviewer (2026)
  • OSDI Artifact Evaluation Committee (OSDI ‘25, OSDI ‘26), ATC ‘25 Artifact Evaluation Committee

Other Services:

  • Berkeley EECS Graduate Student Assembly, Co-President for CS (AY 2026 - 2027)
  • Berkeley NEXT Meetup Organizer (2026 - present)
  • Berkeley Faculty Hiring Student Committee Chair, EECS Graduate Student Assembly, EECS Grads Matter Committee Member (2025)
  • Berkeley Programing System Seminar (PS2) Organizer, Summer Edition on Research Agents (2025)
  • Berkeley Graduate Woman of Engineering, Career and Professional Develoment Committee (2024, 2025)
  • Berkeley CSGE Seminar Chair (2024 - present)
  • Berkeley Faculty Hiring Student Committee Member (2024)

Hobbies/interests

  • 🧗‍♀️ I love bouldering: I am a solid V2 climber, my current goal is to climb a V5 route!
  • ☕️ My favourite drinks are Iced Americano and Iced Matcha 🍵. I also like making latte art, pattern progress: blob (100%), solid heart (85%), rosetta (20%), ??? (100%).
  • 🐶 I love my dog, her name is Paofu. She holds all six level of certifications on behavours and tricks 🎓
  • I use emoji exccessively.

This website is last updated: Aug 2026