Senior Product Manager

Responsibilities

• Own the roadmap for data validation across the invoice lifecycle — capture, extraction, coding, and matching — with a clear goal of raising straight-through processing while protecting accuracy.
• Define how the platform reasons about confidence: when to auto-process, when to flag for review, and how to present exceptions so a human can resolve them in seconds.
• Partner with ML and data science on extraction and GL-inference model performance — framing the problems, defining evaluation metrics, and prioritizing the data and feedback loops that improve them over time.
• Design the human-in-the-loop validation experience so that every correction an AP team makes feeds back into a smarter model.
• Own the metrics for data quality — field- and line-item-level accuracy, exception rate, correction volume, and time-to-resolution — and use them to drive the roadmap.
• Get deep on the messy realities of real invoices: multi-format documents, non-PO spend, complex multi-location chart structures, and long-tail vendors.
• Work with customer-facing teams to understand where validation breaks down in the field, and close the gap.

• 5+ years in product management, with meaningful experience on data, ML, or automation products.
• Comfort working alongside data scientists and ML engineers — you can frame an extraction or classification problem, reason about precision/recall trade-offs, and define what "good" means in metrics.
• A strong sense for human-in-the-loop product design: how to balance automation with human judgment, and how to make review fast.
• Analytical rigor — you instrument, measure, and let data settle debates.
• Experience with B2B SaaS; fintech, AP/AR, document processing, or accounting-adjacent products a strong plus.
• Familiarity with OCR/LLM extraction pipelines, intelligent document processing, or confidence-scored automation.
• Domain knowledge in accounts payable, GL coding, or finance operations.
• Experience building feedback loops that convert user corrections into model improvements.