Staff Plus Software Engineer, Anthropic

• Are proficient in Python and have experience with ML frameworks like PyTorch, TensorFlow, or JAX
• Understand distributed systems principles and have built systems that handle high-throughput, low-latency workloads
• Have built automated or self-service deployment pipelines and eval infrastructure allowing researchers to roll out classifiers and models independently
• Have implemented A/B testing frameworks and experimentation infrastructure for ML systems
• Are results-oriented, with a bias towards reliability and impact in safety-critical systems
• Enjoy collaborating with researchers and translating cutting-edge research into production systems
• Care deeply about AI safety and the societal impacts of your work
• We encourage you to apply even if you do not believe you meet every single qualification. Not all strong candidates will meet every single qualification as listed

Desirable

• The Safeguards ML Inference Path team designs, builds, and operates the production infrastructure that powers Claude's ML based safety systems. We collaborate closely with safety researchers and inference engineers to bring new classifiers and novel classes of ML defenses to production
• We own the research → production transfer of new safety technologies that is on the critical path for every Claude model launch. And we build for scale: serving thousands of ML classifiers, for all requests on the token generation path, and for every platform Claude runs on -- 1P, Bedrock, Vertex, and beyond
• We’re growing the team and looking for engineers who have deep expertise in productionizing ML systems. You'll work at the intersection of machine learning, large-scale distributed systems, and AI safety, developing the platforms and tools that enable our safeguards to operate reliably at scale. And your tooling and infrastructure will be used for every model launch, which are becoming more complex, and more frequent
• Design and build scalable ML infrastructure to support real-time safety deployments across our classifier and model ecosystem
• Build monitoring and observability tools to track classifier performance, data quality, and system health for safety-critical applications
• Collaborate with research teams to productionize safety research, translating experimental safety techniques into robust, scalable systems
• Optimize inference latency and throughput for real-time safety evaluations while maintaining high reliability standards
• Implement automated testing, deployment, and rollback systems for ML models in production safety applications
• Partner with Safeguards, Security, and Alignment teams to understand requirements and deliver infrastructure that meets safety and production needs
• Contribute to the development of internal tools and frameworks that accelerate safety research and deployment