← fivetran / Staff Product Manager, Developer Experience
brief / art_D5ytMi2yN6I
role
model
anthropic/claude-sonnet-4.6
created
2026-07-02T21:20
Company snapshot
Fivetran is a leading ELT data integration platform that automates data movement from hundreds of sources into cloud warehouses, positioning itself as the reliability layer of the modern data stack. The company is valued at over $5.6 billion and has recently merged with dbt Labs, combining automated data movement (Fivetran) with analytics engineering and SQL transformation (dbt) to deliver an end-to-end data infrastructure layer. This merger is actively in progress — the role sits on the dbt Labs product side, specifically owning the dbt Cloud developer experience surfaces. Fivetran/dbt Labs serves thousands of organizations including large enterprises, with a strong community of analytics engineers and data practitioners. Engineering reputation is generally positive in the data community; dbt in particular has a strong open-source following and developer-centric culture, though specific internal engineering details are not publicly confirmed.
Team stack
dbt Cloud frontend: likely TypeScript/React for Studio IDE and VS Code Extension. Backend/compilation: Rust or Go likely for the Fusion platform (based on JD references to 'compilation, intelligence, and platform capabilities' and dbt's known Rust-based Fusion rewrite signals in the community — uncertain, inferred from public signals). CLI tooling: Python (dbt-core is Python-based). IDE infrastructure: LSP (Language Server Protocol) explicitly mentioned in JD, tree-sitter for syntax parsing (JD-mentioned). AI/Developer Agent: LLM orchestration layer, likely OpenAI/Anthropic APIs with context management, agentic workflow tooling. Data warehouses integrated: Snowflake, BigQuery, Databricks, Redshift (JD-confirmed). Version control: Git-based workflows (GitOps patterns standard in dbt). Observability/telemetry: likely internal tooling for developer engagement metrics (activation, session depth). The Fusion team appears to be a platform/compiler team whose capabilities are surfaced through the developer-facing UX — boundary between platform and product is a key architectural concern per the JD.
Likely questions (10)
| area | question | why |
|---|---|---|
| domain | Walk us through how you would define the product strategy for the Developer Agent — how do you differentiate an AI coding assistant for data practitioners versus general-purpose tools like Copilot or Cursor? | The JD explicitly calls out Developer Agent as a core ownership area and asks for strong opinions on AI coding assistants; they want someone who has studied Copilot/Cursor and can articulate a differentiated vision for the dbt context. |
| system_design | How would you design the context management layer for an AI coding assistant embedded in an IDE like dbt Cloud Studio — what context should the agent have, how do you manage token budgets, and how do you handle multi-file or multi-model dbt project awareness? | JD highlights 'context-aware suggestions' and 'agentic workflows' as core Developer Agent requirements; the Fusion team owns compilation/intelligence, so the PM must reason about the platform-product boundary for context. |
| system_design | The JD mentions LSP and tree-sitter as IDE infrastructure primitives. How would you use a Language Server Protocol implementation to power intelligent features in the dbt Studio IDE — and how do you decide what belongs in the LSP versus the AI layer? | LSP and tree-sitter are explicitly listed as edge qualifications; the role requires working fluently with engineering on IDE infrastructure decisions. |
| behavioral | Tell me about a time you owned a developer platform or SDK product and had to balance the needs of power users (who want raw capability) against newer practitioners (who need guardrails and guidance). How did you resolve that tension in the roadmap? | JD explicitly frames the Developer Agent challenge as 'lowering the floor for new practitioners while raising the ceiling for experts' — this is a stated core tension the PM must navigate. |
| behavioral | Describe a situation where you had to co-own a roadmap with a platform or infrastructure team (not your direct team). How did you establish shared priorities, API contracts, and coordinated shipping? | The Fusion partnership is described as a 'core' relationship requiring 'shared roadmap visibility, clear API contracts, and coordinated shipping' — this is a named success criterion. |
| coding | You're reviewing a proposal from engineering to implement a new dbt Cloud CLI command. Walk me through how you would write the acceptance criteria and define the developer experience contract — what does 'done' look like for a CLI feature? | Cloud CLI is explicitly a first-class product ownership area; the JD asks for someone who can write crisp specs and reason about developer workflow automation. |
| domain | How would you measure the success of the Developer Agent feature — what metrics would you instrument, and how do you distinguish 'the agent is being used' from 'the agent is actually making developers more productive'? | JD success criteria include 'activation, session depth, feature adoption' trending up; they want a PM who can define meaningful developer experience metrics beyond vanity usage numbers. |
| culture | dbt has a strong open-source community and a culture of developer advocacy. How have you engaged with developer communities in past roles, and how would you use community signals to inform the Studio/CLI/Developer Agent roadmap? | JD calls out 'tight feedback loops with the dbt community' and 'background in open source' as an edge qualifier; dbt's community-first culture is a known company value. |
| domain | Fivetran and dbt Labs are mid-merger. As the Developer Experience PM, how would you think about the opportunity to create a unified developer workflow that spans data movement (Fivetran) and transformation (dbt) — and what are the risks of moving too fast on integration? | The JD explicitly references the merger and the combined mission; a Staff PM is expected to think about the strategic opportunity, not just the tactical roadmap. |
| behavioral | Tell me about a 0-to-1 developer-facing product or feature you shipped. What did you learn from the first users, and what would you do differently? | JD requires demonstrated experience shipping developer tools and conducting user research; the role involves evolving existing surfaces (Studio, CLI) and building net-new experiences (Developer Agent agentic workflows). |
Talking points
- At Intuit, I owned the full developer platform lifecycle for ICE — SDK Starter Kits (Java/Python), a self-service DevPortal with GitOps config, and a playground environment — reducing onboarding from 2–3 weeks to under 24 hours for production. I drove 275% YoY engagement growth to 675M+ engagements in FY23 and scaled throughput from 6K to 50K TPS. That's the same 'developer activation and session depth' success model the dbt Developer Experience role is optimizing for.
- I've built production AI agent orchestration from scratch: OpenClaw multi-agent framework with gateway protocol and subagent delegation (StreamIO), and a RAG pipeline with multi-provider LLM orchestration, fallback routing, and token budget optimization (Fintellect). I've also built my own aeval evaluation platform with adversarial safety testing, bootstrap confidence intervals, and CI/CD regression gates — so I bring both a builder's intuition and a rigorous evaluation mindset to Developer Agent product decisions.
- I built an RL post-training workbench that benchmarks GRPO, DPO, PPO, and 9 other algorithms across TRL, VeRL, OpenRLHF, and NeMo RL — with live SSE metric streaming and GPU Docker passthrough. This gives me a concrete, hands-on perspective on what 'AI-assisted development tooling' actually requires at the infrastructure level, which directly informs how I'd reason about the Fusion platform-to-Developer Agent boundary.
- I have direct experience with the IDE infrastructure layer the JD calls out: I've shipped an embedded terminal (xterm.js/node-pty with 20+ slash commands) and integrated MCP server tooling that exposes screen capture capabilities to AI coding assistants inside a desktop IDE (StreamIO). I've also studied LSP and tree-sitter in the context of building intelligent editor features — I can engage with engineering on IDE primitives, not just at the product surface.
- I've conducted enterprise-wide developer research at scale — a 9-language Service Language Assessment at Intuit presented to the CTO, plus ongoing developer feedback loops across ~20 mobile apps and 30+ SKUs. I know how to translate qualitative developer pain into prioritized, scoped roadmap work, which is exactly what the dbt community feedback loop requires for Studio, CLI, and Developer Agent.