Staff Product Manager (Compute)

Lambda; United States; Vor Ort; USD 150.000 - 190.000; Vollzeit

Benefits dieser Stelle

Health insurance

401k match

5 sick days and 12 paid holidays

Flexible PTO

Paid parental, medical, and caregiver留

Zusammenfassung

Lambda in the United States is seeking a product manager for the Compute team to shape how customers access compute — GPUs and CPUs — and how capacity is allocated and managed.

You will collaborate across storage, networking, orchestration, and commerce to land a cohesive product, write clear requirements, and measure adoption and utilization with data.

Qualifikationen

• Able to define iterative plans that move an organization from the current state towards the desired outcome.
• Have owned pricing, packaging, or commitment terms for an infrastructure product.
• Have shipped compute primitives that external customers used at scale, like instance types, capacity products, provisioning interfaces, or placement and isolation controls.
• Can turn ambiguous customer, technical, and commercial inputs into product definition that multiple teams execute against.
• Communicate plainly, write well, and make decisions easier for people who do not all share the same context.
• Can reason about utilization and capacity economics, and make the call when customer flexibility and hardware utilization pull against each other.
• Have owned a product where day two behavior mattered as much as launch, including failure handling, maintenance, and what customers are told when something breaks.
• Understand GPU and CPU infrastructure well enough to reason with engineers about launch paths, hardware failure modes, placement, and performance.

Aufgaben

• Find the Real Problem: get close to customers, deals, escalations, and support on On-Demand GPU Instances and 1-Click Clusters.
• Decide What Matters: rank the work and sequence it, and be honest about what Lambda is not doing this year and why.
• Write the Definition: produce requirements, user stories, and acceptance criteria precise enough that engineering builds from them without a translation layer.
• Price and Package It: decide how your part of compute is priced, packaged, and committed to across NVIDIA GPU generations.
• Deliver With Engineering: work through the build with compute and control plane teams, and keep scope honest against the date.
• Land the Launch: set launch criteria covering operational readiness, and get documentation, pricing, sales, and support in place before it goes live.

Kenntnisse

PM experience

Compute platform

GPU/CPU infra

Performance isolation

Cross-team collaboration

Jobbeschreibung

• As a product manager on the Compute team, you will own a meaningful part of defining the future of how we provide compute to our customers
• The domain covers the shape of what customers rent, meaning instance families, sizes, and tenancy, across both GPUs and central processing units (CPUs)
• It covers how customers get capacity and hold onto it
• It covers what they build production automation against, and how fast and predictably that behaves
• It covers what happens once the workload is running, which means placement, performance isolation, hardware failure, maintenance, and the signals customers need to run their own operations
• You will work across these areas depending on which customer needs are burning hottest
• Most of this is the feature set a mature compute cloud already has and Lambda does not have yet
• A neocloud is not a hyperscaler, though, and copying that list wholesale is the wrong instinct
• Part of the job is judgment about which of those capabilities matter here and which carry cost we should not pay
• The other part is deciding where serving AI workloads well means building something the hyperscalers never needed
• Great product managers at Lambda are defined by three things: insight, influence, and execution
• Insight means you look at the data, determine what it means for customers and business, and then figure out what to do about it
• But, a great idea doesn’t mean anything in a vacuum
• That is where influence comes in
• Influence means you take that idea and get others to want to buy into it; you win over engineers, designers, executives, and partners without relying on authority
• But a great idea that everyone is excited about doesn’t matter unless it is delivered to customers
• Execution means you work with the right people to get the idea launched, then measure and iterate
• We hire product managers who learn new domains fast and reason rigorously from evidence
• Deep compute platform experience at a hyperscaler or neocloud, across both GPUs and CPUs, is highly desired
• If you have built compute primitives at a cloud provider and want to do it again somewhere the answers are not settled, we’d love to hear from you
• We value diverse backgrounds, experiences, and skills, and we are excited to hear from candidates who can bring unique perspectives to our team
• Find the Real Problem: Get close to customers, deals, escalations, and support on On-Demand GPU Instances and 1-Click Clusters, and work out what is actually blocking them rather than what they asked for
• Decide What Matters: Rank the work and sequence it, and be honest about what Lambda is not doing this year and why
• Write the Definition: Produce requirements, user stories, and acceptance criteria precise enough that engineering builds from them without a translation layer
• Price and Package It: Decide how your part of compute is priced, packaged, and committed to across NVIDIA GPU generations, and how customers compare the options
• Deliver With Engineering: Work through the build with the compute and control plane teams, make the tradeoff calls that come up mid-flight, and keep scope honest against the date
• Land the Launch: Set launch criteria that cover the operational readiness an infrastructure product needs, and get documentation, pricing, sales, and support in place before it goes live rather than after
• Measure and Iterate: Define what success looks like before launch, then go find out whether it happened, using adoption and utilization rather than opinion
• Make the Call Stick: Take a position on contested tradeoffs, write it down well enough that people can disagree with it precisely, and keep owning the decision after it is made
• Work as a Group: Partner with the other product managers on compute and with the teams that own storage, networking, orchestration, and commerce, so the pieces land as one product

Benefits

• Health, dental vision
• 401k match
• 5 sick days and 12 paid holidays
• Flexible PTO
• Paid parental, medical, and caregiver leave

• Able to define iterative plans that move an organization from the current state towards the desired outcome
• Have owned pricing, packaging, or commitment terms for an infrastructure product
• Have shipped compute primitives that external customers used at scale, like instance types, capacity products, provisioning interfaces, or placement and isolation controls
• Can turn ambiguous customer, technical, and commercial inputs into product definition that multiple teams execute against
• Communicate plainly, write well, and make decisions easier for people who do not all share the same context
• Can reason about utilization and capacity economics, and make the call when customer flexibility and hardware utilization pull against each other
• Can tell which parts of a hyperscaler playbook transfer to an AI cloud and which are cost without benefit
• Have owned a product where day two behavior mattered as much as launch, including failure handling, maintenance, and what customers are told when something breaks
• Understand GPU and CPU infrastructure well enough to reason with engineers about launch paths, hardware failure modes, placement, and performance
• Have 7+ years of product management experience, including time on a compute platform at a hyperscaler, neocloud, or comparable cloud provider
• Experience with performance isolation and placement on shared hardware, including what a provider can honestly guarantee
• Experience with bare metal or dedicated host products, including isolation between tenants
• Experience with distributed training or large-scale inference workloads and what they demand of the compute layer
• Experience with hardware failure handling across a large hardware footprint, like spare pools, node replacement, or fault containment
• Experience with the surfaces customers automate against, like public APIs, infrastructure as code providers, machine images, or snapshots
• Experience with interruptible or preemptible capacity, or with reservations and committed use products