Sr. Product Manager, Technical, Amazon Selection and Catalog Systems

Amazon is seeking a Senior Product Manager - Technical to own the vision and roadmap for MARS (Model Accelerator and Recommender Service), our hardware-agnostic platform for optimizing, benchmarking, and containerizing large language models across heterogeneous compute (Inferentia2, Trainium, and GPU). MARS automatically tunes and recommends optimal hosting configurations, delivering substantial inference cost reduction, throughput gains, and accelerator-utilization improvements for teams across the organization-from model hosting and catalog enrichment to search ranking.

In this role, you will work backwards from internal customers to define a multi-year product strategy that scales MARS into a shared, self-service optimization platform. You will partner closely with science and engineering to expand the platform"s capabilities-new hardware backends, quantization and distillation techniques, and automated benchmarking-while establishing the operating model, intake, and prioritization framework needed to grow adoption. You will own key business metrics such as cost per million tokens, accelerator utilization, and realized IMR savings, and communicate progress and trade-offs to senior leadership.

The ideal candidate is a technical product leader who thrives at the intersection of ML infrastructure, cost optimization, and cross-org influence, and who can turn ambiguous, high-leverage problems into measurable outcomes.

Key job responsibilities

As a Senior Product Manager - Technical for MARS (Model Accelerator and Recommender Service), you will own the product vision, strategy, and roadmap for Amazon"s hardware-agnostic model optimization and benchmarking platform. MARS automatically optimizes, containerizes, and recommends optimal hosting configurations for LLMs across heterogeneous compute (Inferentia2, Trainium, and GPU), driving significant inference cost reduction and throughput gains for teams across the organization.

You will:

Define and drive the multi-year MARS product roadmap, translating cost-optimization and capacity goals into prioritized, measurable deliverables.

Partner with science and engineering to expand MARS capabilities-new hardware backends, quantization and distillation techniques, and automated benchmarking.

Work backwards from internal customers (model hosting, search ranking, catalog enrichment) to identify high-leverage optimization opportunities and unblock adoption.

Establish the MARS operating model, intake process, and prioritization framework to scale the platform as a shared service.

Own key business metrics-cost per million tokens, GPU/accelerator utilization, throughput, and realized IMR savings-and report progress to senior leadership.

Drive cross-org alignment across partner teams and lead the roadmap workshop cadence.

Manage stakeholder communication, technical trade-offs, and go/no-go decisions across the platform lifecycle.

A day in the life

1. Prioritize the MARS backlog and refine requirements with engineering and science partners.

2. Review benchmarking results, optimization performance data, and cost/utilization dashboards to inform decisions.

3. Engage internal customer teams to capture needs, resolve blockers, and drive platform adoption.

4. Lead or attend standups, roadmap workshops, and cross-org syncs; document decisions and action items.

5. Triage incoming optimization requests through the intake process and set expectations on timelines.

6. Write and update product documents (PRFAQs, roadmap artifacts, execution plans).

7. Track KPIs, escalate risks, and communicate progress to stakeholders and leadership.