Senior AI Product Manager, Code

Senior AI Product Manager

Scale has been the leading AI data foundry, helping fuel the most exciting advancements in AI, including frontier model training, enterprise adoption, defense applications, and autonomous vehicles. Our mission is to develop reliable AI systems for the world's most important decisions.

Coding is one of Scale's strongest and fastest-moving domains. We built SWE-Bench Pro, a contamination-resistant benchmark of 1,865 long-horizon software engineering tasks across 41 repositories — including a first-of-its-kind private set drawn from proprietary startup codebases — which cut frontier model scores from over 70% on SWE-Bench Verified to roughly 23%, and became a reference point for how the industry measures coding agents.

We then extended that foundation with SWE Atlas, an evaluation suite spanning Codebase QnA, Test Writing, and Refactoring, which measures the full engineering loop rather than issue resolution alone. And we are among the largest external contributors to the Frontier Bench (Formerly Terminal-Bench) lineage, contributing more tasks than any other single organization to the launch set.

We're looking for a Senior AI Product Manager to own and scale this Coding portfolio from here. In this role, you will define the strategy, roadmap, and operational excellence of Scale's coding data products, RL environments, and agentic coding evaluations. You will work across AI Product Management, ML Researchers, Engineering, Operations, and Go-To-Market teams to turn our benchmark credibility into a durable, revenue-generating product line that leading labs depend on to train the next generation of software engineering agents.

You will serve as the product owner for coding initiatives, driving task design, environment infrastructure, expert contributor quality, customer adoption, and business impact. You will work directly with leading AI labs and enterprise customers, representing Scale as a thought partner in how coding models are trained and measured.

The ideal candidate combines strong product judgment, hands-on technical depth in software engineering, operational rigor, and customer-facing experience, with a passion for turning emerging model capabilities into scalable data products.

You Will

• Own the roadmap and strategy for Scale's Coding portfolio, defining priorities across SFT and preference data, reinforcement learning environments, agentic task suites, and evaluation products.
• Extend the SWE-Bench Pro and SWE Atlas franchises; deciding what comes next as agents saturate current tasks, and converting benchmark authority into training-data and environment revenue.
• Facilitate exploration of the Coding domain and drive alignment among AI-PM, ML, Engineering, Operations, and GTM stakeholders.
• Evaluate, prioritize, and operationalize new coding product proposals, ensuring alignment with customer demand, model capability frontiers, and company strategy.
• Define and manage the end-to-end coding product lifecycle, from ideation and task taxonomy design to pilot, launch, scaling, and sunset decisions.
• Partner with ML researchers and senior software engineers to develop trustworthy task specifications, rubric and grader design, verifiable reward signals, and quality bars for code correctness.
• Drive the roadmap for coding infrastructure;
  Harbor-native container environments, execution sandboxes, reproducible repo images, automated verification, and contributor tooling — to improve scalability, reduce manual effort, and accelerate delivery.
• Establish governance processes for data quality, contamination and leakage prevention, license and IP hygiene, reproducibility, auditability, and release management.
• Grow and steward the expert contributor network of professional software engineers across languages, frameworks, and seniority levels, balancing throughput, cost, and quality.
• Own and grow external partnerships across the coding ecosystem, including open-source benchmark collaborations, enterprise codebase partnerships, and joint releases.
• Work directly with frontier AI labs and enterprise customers to understand where their coding models fail, gather feedback, and influence future…