Prime Intellect and Labelbox hire FDEs to build RL environments and verifiers, a new branch of the role
An FDE Times analysis (Nov 20) looks at Prime Intellect and Labelbox postings for forward deployed engineers whose main job is turning customer workflows into RL environments, eval harnesses and verifiers for training agents.
FDE Times looked at two recent postings that point to a new kind of forward deployed work. Labelbox is hiring a "Forward Deployed Engineer, RL Environments" to build isolated, reproducible execution environments that AI agents train in. Prime Intellect's forward-deployed research engineer role centers on designing RL environments, eval harnesses and verifiers tailored to each customer, and it lists intuition for eval design as a core skill.
The piece cites Epoch AI, which sees enterprise workflows as a major growth area for RL environments, and Wing VC, which notes that enterprises usually get this work through forward-deployed engineering. Wing VC also warns that environment companies without research depth risk getting stuck as on-site contract teams that sell hours rather than a product.
Why it matters for FDEs: the classic skill of sitting with a customer and turning their real work into something measurable is now being sold to AI labs as training signal. Engineers who can build a reproducible sandbox, write concrete test cases and grade an agent on the end state, not its wording, have a path into both deployment and model training. The article also suggests asking early who owns a customer's workflow before it becomes someone else's training data.
Source: fdetimes.net