I possess extensive engineering experience across both Machine Learning Systems (MLSys) and Large Language Model Algorithms. My goal is to advance next-generation AGI systems in order to create larger and better models. I am deeply passionate about the latest technologies and actively contribute to the open-source community as a core contributor to several popular open-source AI projects.

EDUCATION BACKGROUND

National University of Singapore

2022.8 - 2024.1

Master of Computer Science

Singapore

Focus on Machine Learning System in HPC-AI lab
Dissertation: Maximizing parallelism in Distribted training for diffusion model

Zhejiang University

2018.9 - 2022.7

B.Eng in Electronic Science and Technology

Hangzhou, China

Shannon elite class of Information Science and Electronic Engineer College
Minor in Intensive training Program of Innovation and Entrepreneeyrship(ITP) in Chu Kechen Honors College
Honors: First Class Scholarship of Zhejiang University, Excellent Graduate of Zhejiang University

Other Honors: Second prize in the National High school Mathematics Competition(2017)

WORK EXPERIENCE

ByteDance Seed

2023.12 - 2025.12

Staff Research Engineer&&Leader of Multimodal Training Framework

Shanghai

As one of the earliest members of the Seed Team, I focused on AI infrastructure and large-scale training systems for LLMs and multimodal foundation models across pre-training and post-training. Starting from the first-generation Seed models, I supported large-scale training on 10,000-GPU clusters. I led a small team to build VeOmni, an open-source multimodal training system, and supported training at the scale of thousands of GPUs. I was deeply involved in the R&D of the core Seed 1.5 to Seed 2.0 model families, including reasoning and multimodal models, as well as the UI-TARS series of GUI agent models.

Projects Highlights

VeOmni: led the development of a PyTorch-native multimodal training system for pre-training and post-training, supporting model initiatives including Seed core models and UI-TARS.
Core model R&D: participated in the development of the core Seed 1.5 to Seed 2.0 model families, covering major reasoning and multimodal model efforts.
UI-TARS series: contributed to the research and system infrastructure behind the UI-TARS family of native GUI agent models.
Open-source systems: contributed to veScale and verl for distributed LLM training and RL post-training.

ByteDance AML

2023.6 - 2023.12

LLMs Research Intern at Seed-Project

Shanghai

Worked on LLM post-training and agent research, with projects directly leading to publications on process reward modeling, SFT data selection, and data-analysis agents.

Process Reward Modeling: Built the full data-processing, training, and evaluation pipeline for step-level reward models, leading to the paper Let's Reward Step by Step: Step-Level Reward Model as the Navigators for Reasoning.
SFT Data Selection: Co-developed DavIR, a model-centric data selection method showing that 6% of Alpaca data can outperform full-dataset training, later published at ACL 2025.
Agent for Data Analysis: Built InfiAgent-DABench, including the benchmark, agent infrastructure, and evaluation pipeline for data-analysis tasks, later published at ICML 2024.

Joined as Employee #15 and worked on large-model systems and open-source products from Seed to Series A.

Core developer of ColossalAI, contributing to heterogeneous memory management, pipeline parallelism, and distributed checkpointing.
Led the development of ColossalChat, including instruction data processing, distributed training, and alignment tuning for Coati 7B/13B.
Led ColoDiffusion, building an efficient diffusion training solution with memory-optimization techniques for large-batch training.
Helped grow ColossalAI from 0 to 20k+ GitHub stars and supported related open-source/community impact.
Technology Stack: Python, C++, CUDA, PyTorch, Ray, ColossalAI, PyTorch Lightning, TensorRT, DeepSpeed, Hugging Face

Participated in the development of SenseTime's early Megatron-style large-model training framework.

Contributed to MindSpore and MindSpore Lite on device-side GPU inference and runtime infrastructure.

PUBLICATION

Ma, Qianli, Zheng, Y., Shi, Z., Zhao, Z., Jia, B., et al. (2025). VeOmni: Scaling Any Modality Model Training with Model-Centric Distributed Recipe Zoo. [Paper]
ByteDance Seed, Chen, J., Fan, T., Liu, X., Lin, Z., et al. (2025). Seed1.5-Thinking: Advancing Superb Reasoning Models with Reinforcement Learning. [Paper]
Qin, Y., Ye, Y., Fang, J., Wang, H., Liang, S., et al. (2025). UI-TARS: Pioneering Automated GUI Interaction with Native Agents. [Paper]
Zhou, H., Liu, T., Ma, Qianli, Zhang, Y., Yuan, J., et al. (2025). DavIR: Data Selection via Implicit Reward for Large Language Models. ACL 2025. [Paper]
Hu, X., Zhao, Z., Wei, S., Chai, Z., Ma, Qianli, et al. (2024). InfiAgent-DABench: Evaluating Agents on Data Analysis Tasks. ICML 2024. [Paper]
Ma, Qianli, Zhou, H., Liu, T., Yuan, J., Liu, P., et al. (2023). Let's reward step by step: Step-Level reward model as the Navigators for Reasoning. [Paper]

KNOWLEDGE & SKILLS

Program Language: Python, C++/C, Golang, Javascript
MLSys Full Stack: ColossalAI, VeOmni, DeepSpeed, Ray, Megatron-LM, verl, vllm, transformers, triton
LLMs Full Stack: Pre-train, SFT, RLHF, RLVR, Vision Language Model, Omni Model, Agentic
Tools:Linux,Vim,shell,Git,Docker,Cmake

CLUBS & ORGANISATIONAL EXPERIENCE

Zhejiang University Internet Society Technology department AI lab

2021.10 - 2022.8

String Program Technology department Member of the machine learning subdepartment

2020.7 - Present

Zhejiang University Electroacoustic Orchestra Drummer of Six o'clock studio band

2018.11 - 2021.2

Qianli Ma

EDUCATION BACKGROUND

WORK EXPERIENCE

PUBLICATION

KNOWLEDGE & SKILLS

CLUBS & ORGANISATIONAL EXPERIENCE