Partly - Principal/Staff Product Manager – Interpreter

• 💻 What you'll own • Automation Rate. The proportion of jobs fully resolved by Interpreter without human intervention. More than any other metric, this is the one that challenges 17 million manual workers — and determines whether Partly scales as infrastructure or plateaus as a tool. • User Experience. How repairers and estimators interact with Interpreter's outputs — damage capture, guided workflows, how results are presented and acted on. The surface where model capability becomes customer value, and where friction directly suppresses the automation rate above it. • Diagram Quality. The accuracy, completeness, and usefulness of the interactive diagrams and structured parts views — built in close collaboration with Data Ops, who own the internal annotation loop that underpins them. Much of what makes Interpreter irreplaceable to any general-purpose AI lives here. • Parts Verification & Completeness. The rate at which Interpreter correctly validates parts and produces complete, job-ready parts lists across vehicle types, markets, and estimation inputs. Every downstream workflow is only as reliable as these two numbers. • Model Performance via the Customer Feedback Loop. The volume and quality of correction signals captured from real jobs, and the rate at which those signals translate into accuracy gains across markets and vehicle populations. This is how Interpreter compounds: every job makes the next one better. • Loading Performance. User-perceived speed from VIN or job submission to actionable output, measured by p50/p95 response times and user-reported friction attributable to latency. In a workflow measured in minutes, every second is felt. • Non-negotiables • Proven experience as a product manager in a data-intensive or AI/ML product environment, with direct ownership of model-dependent user-facing products. • Deep understanding of human-in-the-loop product design: how to structure review workflows, communicate model confidence, and close the feedback loop between user corrections and model improvement. • Strong analytical capability — able to define, instrument, and reason from product metrics, and to distinguish signal from noise in model performance data. • Experience working across technical and non-technical stakeholders, including close collaboration with engineering, data, and operations teams without owning them. • Track record of driving measurable quality or accuracy improvements in a complex, domain-specific product. • You build rapid prototypes yourself — with AI-assisted tools, LLM-assisted development, or code — as a first instinct for testing ideas. • Excellent judgment under ambiguity. You commit, you course-correct, you own the outcome. • 5+ years owning complex B2B software products end-to-end, with clear accountability for outcomes — not just delivery. • Strong signals we look for • Familiarity with the automotive, collision repair, or parts supply chain industry — or demonstrated ability to develop deep domain fluency quickly in a technical vertical. • Experience building products that serve a multi-sided marketplace, with an understanding of how value propositions differ across customer segments. • Exposure to annotation pipelines, training data quality, or ML evaluation frameworks — enough to have informed conversations with data and model teams and understand how product decisions affect model outcomes. • Experience scaling a product across multiple markets with meaningfully different inputs, workflows, or data characteristics. • Comfort thinking in feedback loops, evals, and system-level instrumentation — not just user stories. • Founder background, or early employee at a startup that scaled materially. • You've killed or radically changed something you personally believed in, and can explain why. • Domain knowledge in automotive aftermarket, collision repair, or insurance claims. • Experience with computer vision, multi-modal AI, or structured data extraction products. • Experience building or operating products with meaningful agentic or automation components. • Experience scaling products across multiple geographies or customer segments. • 🧭 How you'll think and operate • "I built a test version of the confidence UI this afternoon. Three repairers have already used it. Here's what I learned." • "The correction rate on this vehicle cohort is an outlier. That matters more than anything in the backlog right now." • "Here's what the data shows. Here's what I think it means. Here's what I'm doing about it. Here's when I'll be wrong." Speed, ownership, and judgment are the baseline. What differentiates the best candidates is the ability to operate fluidly across the human side — repairers, suppliers, insurers — and the technical one — model quality, data pipelines, evaluation frameworks — and to shrink the distance between intent and outcome.