Anthropic is seeking a Research Engineer for its Code RL team to advance AI models' ability to write, edit, test, debug, and ship real software. The role blends research and engineering, involving design of RL environments, building reward signals, running training experiments, diagnosing model performance, and improving pipeline speed and reliability. Candidates should have strong software engineering and Python expertise, with experience in reinforcement learning, PyTorch, distributed training, and performance optimization.
Research Engineer, Code RL (Reinforcement Learning) at Anthropic
Hybrid - San Francisco, CA; New York City, NY
More jobs at AnthropicSalary
USD 500,000 - 850,000
Requirements
Skills
- Strong software engineering skills and deep Python expertise, including async/concurrent programming
- Comfortable owning systems end-to-end and debugging across the stack
- Ability to balance research exploration with engineering implementation and engage rigorously in experimental design and interpreting results
- Care about code quality, testing, and performance
- Passion for the impact of AI and commitment to developing safe and beneficial systems
- Experience with reinforcement learning, RLHF, post-training, or LLM finetuning
- Experience building coding agents, code-execution sandboxes, evaluation harnesses, verifiers, or developer tooling
- Background in program analysis, testing, verification, compilers, or formal methods
- Experience with PyTorch and large-scale distributed training, performance profiling, and optimization of ML systems
- CUDA, GPU, or TPU kernel experience and accelerator-performance intuition
- Experience with virtualization and sandboxed code execution environments
Responsibilities
- Design reinforcement learning environments and coding tasks
- Build reward signals and verifiers that capture what "good code" means
- Run training experiments on frontier models
- Diagnose why a model does (or does not) improve at software engineering work
- Improve the speed and reliability of pipelines to iterate quickly
- Collaborate with alignment and frontier red teams to ensure systems are capable and safe
- Partner with applied production training teams to bring research innovations into deployed models
Technologies
PythonAsync/concurrent programmingReinforcement LearningRLHFLLM finetuningCoding agentsCode-execution sandboxesEvaluation harnessesVerifiersDeveloper toolingProgram analysisTestingVerificationCompilersFormal methodsPyTorchLarge-scale distributed trainingPerformance profiling and optimizationCUDAGPU kernelsTPU kernelsAccelerator performanceVirtualizationSandboxed code execution environments
See if your resume is ready for this job
See how our AI can optimize your resume and improve your chances for this role.