Zhitong (Payton) Guo
My work focuses on ML infrastructure and systems for large-scale model training. I am currently a Software Engineer at Meta Superintelligence Labs, on the Kernels & Optimizations team. I'm particularly interested in Computer Use Agents and agent harnesses — building the systems that enable LLM-powered agents to autonomously interact with software environments and complete real-world tasks.
Previously, I interned at Meta on YaRN positional encoding for context length extension, at Alibaba Cloud on distributed computing, and conducted research on LLM agents at Carnegie Mellon University and dialogue systems at Columbia University. I'm driven by building reliable infrastructure that makes AI agents practical at scale.
News
- [2025/12]Our paper TheAgentCompany was accepted at NeurIPS 2025!
- [2025/02]Joined Meta full-time as a Software Engineer on the GenAI Llama Research Platform team!
- [2024/12]Graduated from Carnegie Mellon University with M.S. in Intelligent Information Systems.
- [2024/08]Wrapped up an exciting summer internship at Meta, working on YaRN positional encoding for LLM context extension.
Education
Carnegie Mellon University
Aug 2023 — Dec 2024M.S. in Intelligent Information Systems
School of Computer Science, Language Technologies Institute
New York University
Sep 2019 — May 2023B.S. in Computer Science & Data Science
Courant Institute of Mathematical Sciences
Industry Experience
Software Engineer, Kernels & Optimizations, Superintelligence Labs
Feb 2025 — Present- Building ML infrastructure for large-scale model training across GPU architectures.
- Developing end-to-end test infrastructure (Granary) for nightly benchmark validation.
- Enabling training workloads on next-generation hardware (GB200/GB300).
Software Engineer Intern, Llama Model Training
May 2024 — Aug 2024- Implemented YaRN positional encoding for context length extension during supervised fine-tuning.
Software Engineer Intern, Distributed Computing
May 2022 — Aug 2022- Developed cloud-native solutions for distributed computing infrastructure and internal tooling.
Research Experience
Research Assistant
- Advised by Prof. Graham Neubig.
- Worked on LLM-based agents for software engineering tasks, contributing to the TheAgentCompany benchmark for evaluating AI agents in realistic workplace scenarios.
Research Assistant
- Advised by Prof. Zhou Yu.
- Researched dialogue systems and conversational AI, focusing on task-oriented dialogue and natural language understanding.
Research Assistant
- Conducted research on machine learning applications for domain-specific knowledge-guided dialogue systems.
Publications
TheAgentCompany: Benchmarking LLM Agents on Consequential Real World Tasks
Frank F. Xu, Yufan Song, Boxuan Li, Yuxuan Tang, Kritanjali Jain, Mengxue Bao, Zhitong Guo, et al.
NeurIPS 2025, 2025 [paper]
DAsk: A Domain-Specific Knowledge-Guided Dialogue Dataset for Multi-Professional Scenarios
Zhitong Guo, et al.
Preprint, 2024
Machine Learning Approaches for Photovoltaic Performance Prediction
Zhitong Guo, et al.
Conference Paper, 2023