Senior Research and Development Software Engineer (Fivetran AI)

About this role

• Fivetran and dbt are building the open data infrastructure that powers AI agents you can trust
• Fivetran is looking for a Senior R&D Software Engineer to join our fast-growing Fivetran AI team
• Your data stack was built for humans — but agents are the new primary data consumers, and they have fundamentally different requirements
• Agents can’t intuit context; it must be explicitly codified, governed, and traceable
• We’re building the governed context layer that solves this problem: Agents Schema, an open standard for storing agent-ready context directly in the customer’s own data warehouse, and Context Builder, the managed service that keeps it filled and fresh
• This role goes well beyond standard engineering
• You’ll research emerging techniques in the fast-moving AI landscape and bring real product and market understanding to decide which ideas are worth pursuing — and then you’ll take what you’ve learned and ship it as production software
• We’re looking for a true generalist who is willing and able to wear whatever hat the moment calls for: prototyping a new retrieval technique one week, hardening a backend service the next, then doing SRE or QA work when the team needs it
• Fivetran AI operates like a startup within Fivetran, and we need engineers who thrive on that range rather than staying in one lane
• Builds Infrastructure Agents Can Trust — join our mission to deliver the governed context layer that AI agents depend on: accurate semantic definitions, traceable lineage, data contracts, and auditable history baked in from the start
• Embraces Open Standards — help build portable, interoperable data infrastructure: Agents Schema, open formats (Iceberg, Delta Lake), MCP-native interfaces, and connector skills that work with any model and any compute
• Scales Without Breaking — work to make Fivetran AI efficient at agent scale, where unit costs deflate as volume grows and context retrieval is fast, accurate, and cost-controlled
• We emphasize using no-non...