Research Engineer

Research EngineerSan Francisco, CA, USA • Full-timeAbout ThirdLayerThirdLayer is solving one of the hardest problems in deploying agents: AI models are generic, but people's work and processes are specific. Our product, Dex, embeds within your computer and team, ingesting a continuous stream of data across your browser and work apps to understand how you actually operate. So when you delegate tasks, agents know exactly the context of where information lives and how decisions get made. We're a deeply technical team with a high bar for talent and a shared belief that exceptional people are the foundation of everything we build.The RoleAs a Research Engineer, you'll build the systems that let us post-train models continuously: the training pipelines, environment infrastructure, and evaluation harnesses that turn research into something that runs every day. You'll sit between research and product, taking methods that work in an experiment and making them work across deployments, at scale, without babysitting. That means building the machinery for trace collection, data curation, environment generation, and the RL stack that improves agents over time. When a training run fails at 2am or an eval disagrees with reality, you're the person who figures out why. The role rewards engineers who are rigorous about ML and unusually good at systems, or the reverse.What You'll DoBuild and own the pipelines behind our post-training runs, from data ingestion through deployment.Turn research prototypes into reliable, repeatable systems that run against production data.Build infrastructure for generating, scaling, and versioning training environments and evals.Develop the tooling that feeds the learning loop: searching traces, mining them for signal, labeling and curating data.Build the instrumentation we use to inspect, debug, and understand training runs.Work with researchers to co-design methods and systems, and with product engineers to ship what comes out.What We're Looking ForExcellent software engineering fundamentals paired with genuine ML depth. You've trained models yourself, end to end.Hands-on experience with distributed training, high-throughput data systems, or running large fleets of experiments.Fluency in Python and PyTorch, JAX, or similar.You can take a paper or a rough prototype and turn it into something shippable.Prior work on RL environments, evals, or data quality systems is a strong plus.Organized under load. You keep several workstreams coherent without dropping rigor.Bias toward ownership: you take a system from idea to production and stay responsible for it.