Senior Product Manager, Inference Platform

At NVIDIA, we are building the foundation & blueprints for how AI workloads are served at scale for NVIDIA employees and DSX Partners. As a Product Manager for Inference Platform, you will help define and drive the products and platform capabilities that enable large-scale models serving across a broad portfolio of models, both open source and proprietary. You will work at the intersection of AI research, infrastructure engineering, and real user needs, shaping how inference is delivered reliably, efficiently, and at scale.

This is high-impact role. You will be expected to bring structure to undefined problem spaces, make progress without complete information, and build conviction through deep engagement with users, engineers, and the broader ecosystem of AI Cloud partners. The right candidate is equally comfortable discussing model serving architecture, token economics and writing a clear product brief.

What you will be doing :

• Define product vision and strategy for inference platform capabilities — including APIs, capacity management, cost management, performance and optimization, and model serving infrastructure.
• Translate user needs and infrastructure constraints into clear requirements and prioritized roadmaps.
• Partner closely with engineering, research, and user groups teams to drive execution from concept through launch.
• Be responsible for end-to-end product lifecycle for inference-related products and platform investments.
• Develop deep understanding of the inference ecosystem — model formats, serving frameworks, API formats, and the tradeoffs that matter at scale.
• Drive clarity in ambiguous situations by framing the problem, identifying what is known and unknown, and proposing a path forward.
• Represent the voice of the user and ensure product decisions are grounded in real needs, not assumptions.
• Track and synthesize developments across the inference landscape: open source model releases, serving frameworks, competitive dynamics, and emerging use cases.

…