Machine Learning Manager - Catalog Duplicates

Manager of Machine Learning Science

Candidates for this position are preferred to be based in Boston, MA and will be expected to comply with their team's hybrid work schedule requirements. Our team's are in office Tuesday, Wednesday, Thursday and remote on Monday and Fridays.

Wayfair is an online retail platform with the mission to enable everyone to live in a home they love. Delivering on that mission at global scale requires a high-quality, trustworthy product catalog that customers and internal systems can rely on.

The Catalog Health Science organization builds the machine learning systems that power catalog quality end-to-end: how products enter the catalog, how they are structured, and how they are presented to customers. Within this group, the Duplicates program focuses on one of our most foundational problems: identifying and resolving duplicate and option-variant listings across tens of millions of products, and preventing new duplicates from ever reaching the site.

Our north star is an orchestrated system that Detects, Reviews, and Resolves duplicates end-to-end with minimal human touch, while protecting customer trust, reducing supplier friction, and improving operational efficiency.

We are looking for a Manager of Machine Learning Science to own the science strategy and execution for the Duplicates program across detection, review, and consolidation.

Set Strategy & Direction for the Duplicates Program

• Define and own the ML/AI strategy for product deduplication across the product lifecycle, and aligned with broader ML/AI team roadmaps
• Translate an ambitious north star (automated end-to-end deduplication) into a sequenced set of deliverables that balance impact, risk, and technical complexity.
• Partner with product, engineering, analytics, and catalog operations leaders to prioritize work, shape problem definitions, and align on success metrics for duplicate prevention, backlog reduction, and catalog quality.

Lead ML System Design Across Detect → Review → Resolve

• Oversee the design and evolution of ML models that detect exact and near-duplicate relationships at scale, using a combination of representation learning, similarity search, graph-based methods, and large language models.
• Work with partners to modernize the Review layer – including GenAI-augmented auto-review, human-in-the-loop queues, and QA workflows – so that detection output is converted into high-quality decisions efficiently.
• Collaborate closely on the Resolution/Consolidation layer to ensure that confirmed duplicates are merged or blocked automatically wherever possible, with clear contracts between science, product, and tooling.
• Define and refine measurement frameworks (e.g., detection precision/recall, review accuracy, resolution rate, time-to-resolution, net catalog reduction, and customer/supplier outcomes) and ensure they are used to steer the roadmap.

Build, Lead, and Develop a High-Performing Science Team

• Manage and grow a team of Machine Learning Scientists working across the Duplicates work-stream, spanning model development, experimentation, and productionization.
• Provide hands-on technical leadership: review project proposals, model designs, experiment plans, and code; step into the details when the team is tackling particularly complex or high-risk problems.
• Coach scientists on end-to-end ownership – from problem scoping and stakeholder communication through launch, monitoring, and iteration – raising the bar for scientific rigor and business impact.
• Partner with recruiting and other Catalog Science leaders to hire, onboard, and develop diverse talent at multiple levels.

Drive Cross-Functional Execution and Change

• Act as a primary science point-of-contact for Duplicates across Catalog, Merchandising, Operations, and Partner teams; proactively communicate progress, risks, and trade-offs.
• Work with engineering counterparts to ensure that model and data architectures are robust, observable, and cost-efficient, and that platform investments (feature pipelines, training/inference infrastructure, evaluation tooling) unlock reuse across deduplication use cases.
• Collaborate with catalog operations and vendor partners to…