Principal Product Manager, Augmented Memory Grid (AMG)

About the role

WEKA is looking for a Product Manager to own the roadmap and go-to-market for Augmented Memory Grid (AMG), part of the NeuralMesh platform. This is a deeply technical PM role sitting at the intersection of AI inference infrastructure, high-performance networking, and enterprise storage. You will work directly with engineering, GPU/inference partners (NVIDIA, hyperscalers, GPU clouds), and enterprise customers running large-scale LLM inference to define what AMG needs to do next.

What you’ll do

• Own the AMG product roadmap: KV-cache/prefix-cache offload, memory tiering, and integration with inference engines and orchestration layers (vLLM, NVIDIA Triton/TensorRT-LLM/NIM, Kubernetes-based serving).
• Partner with engineering to define architecture trade-offs across GPU memory, networking (RDMA, GPUDirect, NVMe-oF), and distributed storage — translating inference performance bottlenecks (time-to-first-token, throughput, context length) into product requirements.
• Work directly with enterprise customers and GPU cloud partners: Nebius, CoreWeave, TogetherAI, etc., running production inference workloads to gather requirements, validate benchmarks, and prioritize features that reduce cost-per-token and improve SLAs at scale.
• Partner with NVIDIA and other silicon/inference-stack partners on joint roadmap and certification work.
• Define and track benchmarks (TTFT, throughput, cache hit rate) that demonstrate AMG's value versus standard GPU-memory-only inference.
• Support sales and field teams with technical positioning, competitive differentiation, and enterprise deal support.

Must-have qualifications

• Inference ecosystem depth: hands-on product or engineering experience with LLM inference serving — vLLM, NVIDIA Triton/TensorRT-LLM/NIM, Ray Serve, or comparable — and fluency in concepts like KV-cache, prefix/context caching, quantization, and batching strategies.
• Model & systems familiarity: working knowledge of how modern LLMs are served in production (context windows, multi-tenant serving, GPU scheduling) well enough to translate model-level constraints into infrastructure requirements.
• Networking/infrastructure fluency: comfort with the fundamentals of high-performance networking and distributed systems — RDMA, GPUDirect Storage, NVMe-oF, or equivalent — and how they affect inference performance.
• Enterprise customer experience: track record working directly with large enterprise accounts — requirements gathering, production deployments, SLAs — not solely self-serve/PLG products.
• 10+ years of product management experience, ideally with some portion in infrastructure, ML platforms, or developer-facing technical products.

Nice-to-have

• Prior experience at a GPU cloud, inference platform, or AI infrastructure startup.
• Familiarity with storage systems (parallel/distributed file systems, object storage) in AI/ML pipelines.
• Experience partnering directly with NVIDIA or other accelerator/silicon vendors.