Product Manager, AI Factories Orchestration, Agent Harness

About DriveNets

DriveNets is a leader in high-scale networking software for AI infrastructure and service providers. The company pioneered a disaggregated networking architecture that transforms the economics of large-scale networks while maximizing performance, utilization, and operational efficiency. DriveNets-powered networks are deployed by global leaders, including AT&T and Comcast, supporting more than 30% of total U.S. internet traffic. DriveNets AI Fabric delivers full-stack networking for AI infrastructures, providing the highest-performance, Ethernet-based alternative to InfiniBand. The solution is deployed by hyperscalers, NeoClouds, and enterprises worldwide. With over $1B raised, DriveNets continues to push the boundaries of modern networking infrastructure.

The Role

DriveNets is seeking a Product Manager to build the Agent Harness for AI Factories orchestration – the layer that lets AI agents safely operate AI infrastructure. Operators of large GPU clusters are drowning in telemetry, alarms, and change requests. Your mission is to give them agents that can investigate incidents, explain root cause, propose and execute changes, and plan capacity, with the guardrails that production infrastructure demands.

As the Product Manager for the Agent Harness, you will own the tools, context, guardrails, evaluation, and runtime that connect LLM-based agents to the DriveNets orchestration platform, for both DriveNets-built agents and the agents our customers bring.

You will join the AI Factories Orchestration product team, a dedicated software product group within DriveNets Product Management. This is a software product role: you will own a platform that customers run every day, with APIs, workflows, user experience and integrations at its core. You will partner with hyperscalers, NeoClouds, and enterprises building AI/HPC GPU clusters, and your work will directly determine how fast they can bring GPUs into service and how much useful work they get out of them.

Responsibilities

• Define the Agent Harness: the tool layer (including MCP servers) that exposes orchestration, telemetry, topology, and configuration capabilities to agents in a safe, well-scoped way.
• Own the first-party operations agents – e.g., incident investigation and root cause analysis, change planning and pre-checks, configuration audit, and capacity planning – and define where they run autonomously versus with a human in the loop.
• Define the guardrails: permissions and RBAC for agents, policy enforcement, approval workflows, blast-radius limits, dry-run and rollback, and full audit trails of agent actions.
• Own context engineering: how topology, live telemetry, logs, runbooks, and documentation are retrieved and presented to models to produce accurate, grounded answers.
• Build the evaluation framework – offline eval suites, regression benchmarks on real incident scenarios, and online quality metrics – and make eval results a release gate.
• Make the platform agent-friendly: work with platform teams so APIs, errors, and data models are easy for agents to use, and support open interoperability (MCP, A2A) for customer-owned agents and AIOps tools.
• Stay current with the fast-moving agent ecosystem (agent SDKs, models, open standards) and drive build/partner decisions, including model selection, cost, latency, and on-premise / air-gapped deployment.
• Own the product strategy, roadmap, and backlog for your area, and translate customer problems into clear PRDs, user stories, API specifications, and acceptance criteria for engineering teams working in agile delivery.
• Run continuous discovery with hyperscalers, NeoClouds, and enterprise customers: interview operators and platform teams, map their day-0/1/2 workflows, and turn the patterns into prioritized product requirements.
• Design northbound APIs (REST/gRPC, OpenAPI) and SDKs so the platform can be consumed programmatically by customer portals, cloud control planes, and automation pipelines – not only through a UI.
• Drive integrations with the AI infrastructure ecosystem, such as Kubernetes, Slurm, NVIDIA Base Command Manager / Mission Control, Run:ai, NetBox, ServiceNow, and observability stacks.
• Define and track product success metrics such as time-to-first-job, GPU utilization, mean time to resolve, adoption, and customer satisfaction.
• Lead demos and proof-of-concepts with customers, and support sales, solution architects, and customer success in strategic engagements.
• Create external content – white papers, technical briefs, blogs, conference talks – that balances technical accuracy with clear messaging.

What we need to see

• Mandatory: daily, hands-on use of AI agentic tools (e.g., Claude / Claude Code, Cursor, ChatGPT / Codex, GitHub Copilot) as a core part of how you work – for research, writing PRDs and specs, data analysis, prototyping, and automating your own workflows. Be ready to show us how you use them.
• 6+ years in the technology industry, with at least 3 years as a product manager for technical or developer-facing software, including hands-on experience shipping LLM or agent-based features.
• Strong, hands-on understanding of how LLM-based agents work: tool calling, retrieval (RAG), MCP, and their common failure modes.
• A clear point of view on trust and safety for agents that act on production systems: permissions, approvals, and auditability.
• Proven experience writing clear product requirements for software platforms, with APIs as a first-class part of the product.
• Strong, hands-on understanding of cloud-native software architecture – Kubernetes, containers, microservices, and distributed systems – with a proven ability to drive design trade-offs with engineering.
• Solid understanding of AI infrastructure: GPU cluster architecture, how training and inference workloads use compute, network, and storage, and what drives GPU utilization and job reliability.
• Strong data-driven decision making, with a proven ability to define success metrics and act on product analytics.
• Clear written and verbal communication, with the ability to work with executives, engineers, and customers alike.
• Ability to manage shifting priorities in a fast-paced environment and to work with teams across geographies.
• Travel as needed.

Ways to stand out from the crowd

• Practical experience with LLM evaluation: building eval sets, measuring accuracy and hallucination, and using results as release gates.
• Hands-on experience with agent frameworks and SDKs such as LangGraph, the Claude Agent SDK, OpenAI Agents SDK, or similar, and with agent tracing/observability tools.
• Experience with multi-agent systems, memory, and context engineering over large volumes of live telemetry and logs.
• Background in SRE, NetOps, or AIOps – you have been paged and know what a good root cause analysis looks like.
• Experience deploying LLM solutions in on-premise, sovereign, or air-gapped environments, including open-weight models.
• Ability to prototype: proficient in Python and building a working demo agent yourself.
• Familiarity with data center networking: leaf/spine and Clos topologies, BGP/EVPN-VXLAN, RoCEv2, and AI scale-out / scale-up fabrics (Ethernet, UEC, InfiniBand, NVLink).
• Hands-on experience with infrastructure-as-code and DevOps tooling: Terraform, Ansible, Helm, GitOps, CI/CD.
• Familiarity with observability and telemetry: Prometheus, Grafana, OpenTelemetry, streaming telemetry (gNMI/gRPC).
• Experience at a Tier-1 cloud (AWS, Azure, GCP, OCI), a NeoCloud, or a company building infrastructure management software.
• Experience building AI-assisted features (copilots, natural-language queries, AIOps) into an operations product.

Education

BS/MS in Computer Science, Computer/Electrical Engineering, or another engineering field, or equivalent experience.

If your experience is close but doesn’t fulfill all requirements, please apply. DriveNets is on a mission to build a special company comprised of individuals with different backgrounds, perspectives, and experiences.

More About DriveNets

• Based in Israel with locations in Romania, US, India and Japan as well as extended teams, DriveNets operations cover more than twelve countries. With recognition by industry analysts and through partnerships with market leaders such as AMD, Broadcom, Dell and others, DriveNets is pushing market momentum, delivering the scale and efficiency that modern AI workloads demand. Visit our website: