Our Team 

The CRAFT (Capabilities, Routing, Adaptation & Fine-Tuning) team at CoreWeave is building tools to help agents learn from experience. This is a critical step to make agents reliable enough to perform long tasks autonomously, in the same way human employees are. We’re systematically identifying and solving the major bottlenecks between today’s tech and those future self-improving agents. So far, we’ve: 

• Released ART, the easiest library for getting started with RL. 
• Developed RULER, a general-purpose reward function that works across many diverse tasks. 
• Built Serverless RL, an elegant API that gives RL practitioners full control over their data, environment and reward function while letting them outsource the headaches of managing GPU infrastructure. 

These releases have a theme: we’re systematically tackling each major roadblock to successfully training self-improving agents. Several serious challenges remain. Building simulated environments often requires substantial human labor, and existing training methods are not data efficient enough. We're laser-focused on solving these problems and making self-improvement a reality for agent developers. In startup terms, this is a classic hard-tech bet. Our roadmap involves substantial technical risk; there are still major technical problems we’re facing without a proven solution. However, there is very little market risk. We’ve worked closely with the teams building agents at many of the top AI-native startups as well as large enterprises. If we can build this, everyone will want it. A self improving agent that learns from experience the way a human employee would could quickly capture a large fraction of the total inference market, which is worth tens of billions of dollars today and will be worth hundreds of billions in a few years. 

About the Role 

You have trained LLMs to be SOTA on specific tasks. You have opinions on whether sequence-level or token-level importance ratios are more effective. You probably shared the ScaleRL paper in your group chats, and kicked off a few ablations after you read it. You will be expected to generate and investigate research ideas towards solving the remaining obstacles to continuous learning in production. You will work with the broader CRAFT team to validate these research directions across real customer tasks. We are very GPU rich and are ready to direct an enormous amount of compute at this effort. Beyond your role’s specific qualifications, we’re looking for strong engineers with great taste. The most important qualification by far is that you learn fast and can ship. This role will inevitably involve a lot of learning on the job; we’re building this airplane as we fly it. Engineers on our team touch everything from CUDA kernels to high-performance LLM tracing dashboards, and you will have an opportunity to touch many parts of this stack. Although we operate as part of a larger company, the CRAFT team is small, has a large degree of autonomy and drives our own roadmap and priorities. This is an excellent role for someone looking to find their own company in the future. 

Required Qualifications

• 8+ years of experience in machine learning or applied research, or a PhD with 4+ years of relevant industry experience
• Demonstrated success developing LLM training methods or systems that produce meaningful improvements on real-world tasks
• Deep expertise in LLM post-training, including supervised fine-tuning, reinforcement learning, on-policy distillation, reward modeling, and policy optimization
• Strong research judgment, including the ability to identify high-impact problems, design rigorous experiments, and make decisions from ambiguous results
• Experience taking research ideas from initial hypothesis through implementation, evaluation, and production deployment
• Proven ability to set technical direction, lead complex cross functional initiatives, and mentor other engineers

Preferred Qualifications

• Publications, open source contributions, or other demonstrated research impact in reinforcement learning, LLM post-training, or agent learning
• Deep experience with distributed training, GPU optimization, and large-scale model training systems

Our Stack 

We strive to use the best tool for the job when building and deploying our production services. Sometimes that means writing our own custom code, and often it means leaning on the work of others. As part of building Serverless RL, we depend on the following libraries and frameworks (among many others): 

• Kubernetes 
• Megatron 
• Temporal 
• Postgres 
• FastAPI 

Why CoreWeave? 

We work hard, have fun, and move fast! We’re in an exciting stage of hypergrowth that you will not want to miss out on. We’re not afraid of a little chaos, and we’re constantly learning. Our team cares deeply about how we build our product and how we work together, which is represented through our core values: 

• Be Curious at Your Core 
• Act Like an Owner 
• Empower Employees 
• Deliver Best-in-Class Client Experiences 
• Achieve More Together 

We support and encourage an entrepreneurial outlook and independent thinking. We foster an environment that encourages collaboration and provides the opportunity to develop innovative solutions to complex problems. As we get set for takeoff, the growth opportunities within the organization are constantly expanding. You will be surrounded by some of the best talent in the industry, who will want to learn from you, too. Come join us! 

The base salary range for this role is $185,000 to $275,000. The starting salary will be determined based on job-related knowledge, skills, experience, and market location. We strive for both market alignment and internal equity when determining compensation. In addition to base salary, our total rewards package includes a discretionary bonus, equity awards, and a comprehensive benefits program (all based on eligibility).