Leo Mehr defines forward-deployed engineering at Ramp as product engineering focused on making core and agentic features work for large enterprise customers. He organizes the practice around two principles: always scope the real problemAI customer discovery investigates users, workflows and constraints to find problems where AI can create measurable value. before building, and continually adapt the job so models and agents handle more of the execution pipeline.
Leo Mehr illustrates scoping with two costly failure modes. An urgent integration request may reflect internal sales pressure rather than a customer's actual deadline, while an apparently straightforward mobile feature can waste weeks if the team never verifies which platform the customer uses. Forward-deployed engineers therefore ask who needs the feature, what drives the urgency, whether a workaround exists and whether other customers would benefit before committing engineering effort.
Leo Mehr describes scaling with tokens as applying agents to context gathering, request clarification, specification writing and implementationWorkflow automation uses software or AI to complete a connected sequence of routine steps with less manual effort.. An internal request agent asks multiple rounds of questions until a proposal is ready for a specification, while future work targets the harder middle stages between intake and shipped product. EvalsEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria., rubrics, human feedbackAn AI feedback loop occurs when AI outputs or their consequences become inputs that influence later model or agent behavior. and reliable access to product contextContext engineering designs how relevant instructions, evidence, state and tool information are assembled for a model's task. become central, but forward-deployed engineers retain responsibility for taste and judgment over the final result.
Watch on YouTube
