Ma Xiaoteng / 马肖腾

Overview

Ma Xiaoteng is presented as the co-founder and chief scientist of MineLab, with a Tsinghua automation background and long-running work in reinforcement learning and model post-training.

Current Profile

In the supplied interview, Ma connects three roles: RL researcher, AI entrepreneur, and operator of enterprise post-training systems. His recurring method is to build feedback loops—between action and belief, customer environments and model behavior, or research and commercial usefulness—while keeping evaluation and infrastructure constraints explicit.

Key Characteristics

• Reinforcement-learning researcher with prior large-scale multi-agent training experience.
• Co-founder and chief scientist focused on enterprise post-training and reusable training infrastructure.
• Proponent of exploration, action-based belief revision, and bounded long-termism.
• Advocate for research organizations whose product feedback determines useful technical questions.
• Supporter of a diverse ecosystem of specialized models rather than a single homogeneous model future.

Evidence

Research and systems background

• 锐评国模与 Neo Lab，畅谈 RL 人生与 RLaaS | 对谈 Mind Lab 首席科学家骁腾 describes Ma’s Tsinghua training, RL research, and exposure to large CPU/GPU clusters, Ray, and Kubernetes.

Enterprise post-training

• 锐评国模与 Neo Lab，畅谈 RL 人生与 RLaaS | 对谈 Mind Lab 首席科学家骁腾 attributes to Ma the view that private data, customer benchmarks, and real environments are prerequisites for durable vertical improvement.

Research and life philosophy

• 锐评国模与 Neo Lab，畅谈 RL 人生与 RLaaS | 对谈 Mind Lab 首席科学家骁腾 records his use of exploration/exploitation, policy-value consistency, and discount factors as decision heuristics.

Qualifications

• This profile relies on one structured podcast summary rather than a transcript, CV, publication record, or company filing.
• The source title shortens the guest name to “骁腾,” while the body gives “马肖腾”; the fuller body form is canonical pending stronger evidence.
• Technical performance, project-cost, and company-comparison claims remain attributed to Ma.

What Changed

• Established Ma’s profile around RL research, enterprise post-training, New Lab practice, and model diversity.

Relationships

• MineLab - co-founder and chief scientist.
• Mint - LoRA-centered training service associated with his team.
• Tsinghua University - source-reported education context.
• Enterprise Post-Training Loop - central operating model in the interview.
• Reinforcement-Learning Decision Heuristics - RL concepts he applies to personal and entrepreneurial decisions.