Principal Product Manager
Principal Product Manager at SambaNova Systems who owns the end-to-end model and capability roadmap for its RDU-based AI inference platform. The role prioritizes which LLMs to ship, defines the feature bar (function calling, structured output, long context, caching), stays hands-on testing models in Python, and drives launches with engineering and GTM.
- Compensation
- $190K–$230K / yr
- Location
- San Jose, California, United States
- Work setup
- Not specified
- Seniority
- Principal
- Employment
- Full time
- Experience
- 3+ years
- Category
- Product
- Posted
Skills
Must have
- Bachelor's degree or higher in Computer Science, Engineering, Machine Learning or a related technical field
- 7+ years combined experience in product management and technical roles (ML engineering, solutions engineering, research)
- 3+ years as a PM on AI/ML, developer or infrastructure products
- Strong working knowledge of LLMs: transformer and MoE architectures, inference trade-offs (latency, throughput, context, quantization)
- Knowledge of LLM features such as function calling, structured output and reasoning
- Track record of owning a roadmap and shipping: setting priorities under constraints and driving cross-functional teams to launch
- Comfort with data: building prioritization models, dashboards and scorecards; able to query an API or change a Python script
- Hands-on with models: calling model APIs, running evals and quick experiments, building prototypes (Python, notebooks, lightweight UI tools, AI coding assistants)
- Excellent written and verbal communication; can present a roadmap and defend trade-offs to executives, engineers and customers
Nice to have
- Product experience at an inference provider, model lab, AI cloud or AI hardware company
- Hands-on familiarity with inference stacks (vLLM, SGLang, TensorRT-LLM) and OpenAI/Anthropic-compatible APIs
- Experience designing evals and benchmark methodology for model quality and feature conformance
- Background as an ML or solutions engineer before moving into product
- Experience with enterprise, sovereign AI or public-sector customers
- Active presence in the open-source model community
Full description
SambaNova is a leader in next-generation AI infrastructure, delivering a full-stack inference platform for customers worldwide. At the core of SambaNova's technology is the RDU (Reconfigurable Dataflow Unit) — a chip built on a dataflow architecture rather than the traditional GPU model. Its decode performance is especially strong for agentic workloads like multi-turn agents, code generation, and long-running applications. RDUs are packaged into SambaRack, rack-scale hardware that lets customers deploy state-of-the-art models with better performance, greater energy efficiency, and faster time to value.
We are looking for a technical product manager to own SambaNova's model and capability roadmap end to end. You will be the product owner and the subject-matter expert on models: you understand why a model matters (architecture, benchmarks, adoption, customer pull) and what it takes to ship it well (function calling, structured output, reasoning, long context, caching, speculative decoding and more). Most importantly, you turn that understanding into decisions, commitments and launches. This is a product role for someone deeply technical, not an engineering role with product tasks attached, but you will stay hands-on: testing new models yourself, finding where they shine and where they break, and building quick mock-ups that show the competitive edge fast inference unlocks.
Responsibilities
Own and drive the roadmap
- Model Stack Rank: Own the unified, weighted stack rank of candidate models across customer demand, capability, market adoption, recency, strategic value and engineering effort. Tune the rubric, make and document overrides, and run the regular prioritization review
- Per-Chip Roadmap: Own the plan of models × hardware generation × configuration × performance targets, and keep it grounded in measured platform data rather than estimates
- Commitments and Launches: Turn priorities into dated commitments with engineering, track them to delivery and run model launches with Marketing, Docs and GTM
- Roadmap Communication: Keep the model roadmap and dashboard current as the source of truth for execs, account teams and customers, and present it with confidence both internally and externally
Define the capability bar
- Feature-Completeness Framework: Own the definition of a “fully featured” model across speed, efficiency, model infrastructure, cloud and API features, for example tool calling, JSON schema and constrained decoding, reasoning controls, full context length, prompt caching, speculative decoding, streaming, logprobs and API compatibility (OpenAI Chat, Responses, Anthropic)
- Gap Management: Find capability and conformance gaps from testing and telemetry, set their priority and drive them to closure with engineering