Member of Technical Staff - AI/ML Infrastructure Engineer, RL Post-training
Builds and scales the compute, scheduling, data, and evaluation infrastructure powering RL post-training research on large language models, including LLM inference/training systems, internal tooling, and monitoring/debugging systems. Core stack spans distributed systems on AWS/GCP with Kubernetes, ML frameworks like PyTorch/JAX, and RL training frameworks such as veRL and Ray.
- Compensation
- $200K–$350K / yr
- Location
- San Francisco
- Work setup
- Onsite
- Seniority
- Not specified
- Employment
- Not specified
- Experience
- Not specified
- Category
- Ml / AI
- Posted
Skills
Must have
- Strong software engineering fundamentals and hands-on experience building production-grade LLM inference and training infrastructure (ideally from the ground up)
- Experience building LLM training/inference internals such as transformers and distributed training, and inference libraries like vLLM, SGLang, Megatron
- Experience working on RL training frameworks like Slime, veRL, Ray Train, SkyRL, Miles
- Significant experience with distributed systems principles and hands-on experience with cloud platforms (AWS, GCP) and Kubernetes, building high-throughput, low-latency systems
- Experience with data engineering tools and building robust, scalable data pipelines
- Proficiency in core ML frameworks such as PyTorch or JAX
- Can balance production rigor with fast-moving research pace and communicate infrastructure tradeoffs clearly to non-infra researchers
Full description
About Us
Preference Model is a superintelligence data research company. We build RL environments for training capable, better-aligned superintelligences.
Almost every aspect of how a model behaves is shaped by the reward signals it's trained on. The biggest problem with AIs today is that they don't always do what we intend. Sometimes that's because they aren't capable enough, and sometimes it's because they aren't aligned. Both problems come down to the training objectives.
We're trying to address both of these issues in the most direct, highest leverage way available to us, which today means researching how to build RL environments better, not just for the AIs of today but for the AIs as they will be when they outsmart us.
Over the past year, we've built RL environments for several frontier labs, and we're backed by $16M in seed funding led by a16z. Our founding team has previous experience on Anthropic’s data team building data infrastructure, and datasets behind Claude.
About the Role
Frontier research moves only as fast as its infrastructure permits. Building solid infrastructure is foundational to our mission of pushing self-directed learning as far as it can go.
We are looking for SeniorML Infrastructure Engineers to build the infrastructure and systems that power the frontier of post-training on large language models. This role involves building scalable infrastructure to enable high-throughput systems and shape how our research is run, bringing us closer to models that can train themselves on what they aren't yet good at.
What You Will Do
- Design, build, and scale the compute, scheduling, and data infrastructure that powers post-training research on our in-house RL environments
- Develop and maintain core ML framework primitives and internal tooling that researchers rely on daily, accelerating reproducible experimentation and reducing time from idea to result
- Build evaluation and benchmarking infrastructure, monitoring, logging, and debugging tooling, and automated testing and deployment systems, so failures are caught early and infrastructure stays reliable as it scales
- Partner directly with Research Engineers to translate research needs into infrastructure requirements, and ship fast in response to their feedback
What We are Looking For
- Strong software engineering fundamentals and hands-on experience building production-grade LLM inference and training infrastructure (ideally from the ground up)
- Experience building LLM training/inference internals such as transformers, distributed training, and working on inference libraries like vLLM, SGLang, Megatron