Note: The job is a remote job and is open to candidates in USA. Gigawatt is building an AI platform and common data fabric for utility customers. The Principal Product Manager will own the agent layer, including the SDK, no-code builder, evaluation and observability capabilities, trust and governance features, and the externally shipped platform product experience. The role partners with engineering, design, security, data teams, and utility customers to deliver production-ready capabilities and drive adoption outcomes.
Responsibilities
- Own the AI platform roadmap: Drive strategy and roadmap for the agent infrastructure built on our data fabric, and make the build-versus-buy calls across a fast-moving AI tooling landscape
- Own both build paths: Define the SDK and code-first experience utilities use today, and drive the no-code builder that lets their SMEs ship agents themselves. Decide what gets abstracted, what stays code, and how the two converge over time
- Ship with design partners: Define high-value MVPs, build them with real utility customers, and iterate fast on the friction and edge cases that only show up in someone else’s environment
- Own developer and builder experience: API and SDK design, documentation, sandboxes, templates, onboarding, and the paved path from first agent to production deployment
- Own evaluation and observability: Define how agent quality is measured before and after release: golden sets, regression suites, online evals, and the trace and audit records a utility or regulator can inspect. Give customers visibility into their own agents, not just internal dashboards. Make the build-versus-buy call across LangSmith, Arize, and open standards, and own the decision on where trace data lives
- Design for human-in-the-loop: Define where agents act autonomously, where they pause for approval, and how state, interrupts, and replay work in practice. In utility operations the escalation path is a product decision, not an implementation detail
- Own trust, tenancy, and governance: Explainability, auditability, per-tenant data isolation, customer-controlled retention, least-privilege access, and safe autonomy, especially for agents a non-developer built
- Own the commitments that come with shipping externally: Versioning, backwards compatibility, deprecation policy, SLAs, the support model, and the product input into packaging and pricing
- Execute: Drive with engineering and design to release production-ready capabilities on compressed timelines. Drive the product lifecycle by defining use cases and technical requirements, collaborating on data model and API design, and conducting product testing to ensure key requirements are met
- Own outcomes: Set KPIs (customer agents live in production, time-to-first-agent, share of agents built without code, eval pass rates and regression escapes, agent task success and escalation rate, latency and cost per task, support burden) and manage dependencies across the platform
- Wear many hats: Do whatever it takes to move the platform forward, and take on new areas of ownership as priorities evolve
Skills
- Utility experience is a plus but not a requirement
- 8+ years in product management, including 2+ years on AI or ML systems, and significant time on a platform or product that customers outside your own company built on top of
- You have shipped APIs, SDKs, or an application builder that customers depended on in production. You have lived versioning and deprecation, support escalations, security questionnaires, and the difference between an internal tool and a product someone signs a contract for
- Working fluency in modern AI systems: LLMs, retrieval, tool use, guardrails, and the cost, latency, and reliability tradeoffs that come with them. You can read a trace and tell whether a failure was retrieval, prompt, tool schema, or model, you have designed eval sets, and you have shipped against a quality bar rather than a demo. You can hold your own in an architecture review and write a technical spec engineers respect. You do not need to write production code
- You have shipped software into an environment where someone outside your company audits what it did: financial services, healthcare, energy, or public sector. You have a real point of view on tenant isolation, explainability, data lineage, retention, and least-privilege access, and you treat security and compliance review as a design input rather than a launch blocker
- Remote: prioritizing central and eastern timezones (ideally Chicago, Austin, or Nashville)
- Experience with low-code, configuration-driven, or builder-style products, and a point of view on how to give a business user genuine capability without giving them a way to cause harm. Close to the heart of the role, and the strongest differentiator among otherwise comparable candidates
- Hands-on familiarity with a modern agent framework, ideally LangGraph or LangChain: graph-based orchestration, state and checkpointing, tool and schema design, and multi-agent handoffs
- Experience building on a managed model platform, ideally AWS Bedrock, including model selection tradeoffs, guardrails, throughput and cost management, and VPC, IAM, and data residency constraints
- You will build on our existing data fabric rather than define it, but comfort reasoning about data models, integration across messy enterprise source systems, and streaming versus batch helps you know what the fabric can and cannot answer for an agent
- Utility or energy sector exposure, or a background in software engineering, ML, or data engineering before moving into product
- Experience with Jira, Figma, and standard product management tools
Benefits
- Equity
- Health insurance
- Remote flexibility
- 401k match
Company Overview
Company H1B Sponsorship