Latest AI JobsBrowse Jobs
P

Member of Technical Staff, AI Research

psiBostonTechnical Staff
Apply on company site
Salary
Not listed
Job type
Full-time
Work mode
Remote
Experience
Senior
Posted
7 Oct 2026
Status
Open · checked 9d ago

About this role

Member of Technical Staff, AI Research is a full-time position located in Boston, though remote candidates are considered on a case-by-case basis. This role suits experienced researchers with a quantitative PhD who want to build AI agents and training systems for discovering new physics at scale.

Role Overview

Psi seeks an AI researcher to develop the training systems and agents that learn physics. The position involves working within a startup that aims to industrialize scientific discovery and commercialize transformative physics breakthroughs using artificial superintelligence.

Key Responsibilities

Build and train AI agents and training systems focused on physical reasoning, action space design for scientific tool use, reward structuring for long-horizon tasks, and scalable training infrastructure.
Design evaluation workflows and benchmarks for physics reasoning to separate genuine reasoning from pattern matching and benchmark gaming, while instrumenting agent behavior for interpretability.
Publish findings at top machine learning venues, develop curricula and reward structures for discovery tasks, and iterate on practical implementations.
Partner with domain physicists creating verification harnesses and engineers building training infrastructure to deliver end-to-end working systems.

Requirements

Hold a PhD in computer science, machine learning, physics, mathematics, or a related quantitative field, backed by recent publication history at major venues like NeurIPS, ICML, ICLR, or comparable physics-ML venues.
Possess practical experience constructing agents and training models using reinforcement learning for complex-reasoning domains such as science, mathematics, or code, while deploying RL systems that surpass baseline methods.
Demonstrate proficiency in distributed training and modern machine learning frameworks capable of scaling from single GPUs to clusters.
Bring a foundational background in mathematics or physics that enables intuitive comprehension of scientific tool use and physical reasoning.

Nice to Have

Practical background utilizing modern reinforcement learning algorithms including PPO, SAC, MuZero, search-augmented methods, or multi-agent self-play.
Extensive familiarity with JAX or PyTorch alongside distributed training tools such as Ray, Accelerate, XLA, or similar platforms.
Background deploying agents into games, simulators, scientific tools, or structured benchmark suites.
History of open-source contributions, conference presentations, or adopted research artifacts.

Skills and Tools

PyTorch
JAX
Ray
XLA
Accelerate
PPO
SAC
MuZero
Reinforcement Learning
Distributed Training

Salary and Benefits

Competitive compensation including salary
Benefits
Meaningful early-stage equity

About the Company

Psi is a startup with foundations at institutions including Google, NVIDIA, Harvard, Meta, MIT, Oxford, Johns Hopkins, Cambridge, and the Perimeter Institute. The company builds AI systems designed to discover new physics at scale safely, verifiably, and for broad public benefit. Team members operate with full ownership from specification to shipping and on-call duties, utilizing an AI-native development process featuring agentic coding tools.

How to Apply

Interested candidates should apply on the official company careers page.

Skills and tech stack

PyTorchJAXRayXLAAcceleratePPOSACMuZeroReinforcement LearningMachine Learning

Questions candidates ask

Where is the Member of Technical Staff, AI Research position located?

The role is based in Boston, though remote candidates are considered on a case-by-case basis.

What educational background is required for this role?

A PhD in physics, computer science, machine learning, mathematics, or a related quantitative field is required.

What specific technical expertise is expected?

Candidates need a hands-on track record building agents and training models with reinforcement learning for complex-reasoning domains, alongside proficiency in modern ML frameworks and distributed training.

What are the nice-to-have qualifications?

Preferred qualifications include experience with modern RL algorithms like PPO, SAC, MuZero, multi-agent self-play, and search-augmented methods; fluency in PyTorch or JAX with distributed training via Ray, XLA, or Accelerate; application of agents to simulators, games, or scientific tools; and open-source contributions or conference presentations.

What compensation is offered?

The position provides competitive compensation that includes salary, benefits, and meaningful early-stage equity.

Ready to apply?

The application is on the official psi careers page.

Apply on company site