Braintrust provides an AI observability and evaluation platform that lets engineering and product teams log, monitor, and analyze AI agents and large-scale model interactions.
ABOUT THE COMPANY Braintrust is the agent observability platform. By actively applying intelligence to agent traces and automatically surfacing the most critical patterns, Braintrust gives teams the visibility to understand how agents behave in production and the tools to improve them. Teams at Notion, Stripe, Box, OpenAI, and Cloudflare use Braintrust to trace their agents, find the issues in their observability data, and run evals that tell them how to improve. ABOUT THE ROLE Braintrust is growing quickly across enterprise and self-serve customers, and our internal systems need to scale with us. You will build and own the data infrastructure that will power how our business operations and systems function. This is a hands-on, solo role for now. You won't be joining a team with mature processes already in place. You'll be building the queues, workers, and data models, on top of a warehouse that already has solid foundations (health scoring, ARR movement, engagement facts) but no activation layer. The right person has strong technical depth, understands how GTM teams work, and is excited to build with AI. You should have strong opinions about where agents can automate operational workflows, and the judgment to build the data systems, context, and feedback loops that make those workflows work. You'll partner with Product and Engineering on usage data and event schemas, and work directly with Sales, RevOps, Marketing, and Finance to turn one-off requests into reusable platform primitives (routing rules, scoring models, activation contracts) instead of one-off scripts. WHAT YOU WILL DO - Build and own the customer data platform: a company/event/signal spine that resolves identity across Salesforce, product, and marketing sources into one queryable universe. - Design the queues, workers, and models that power enrichment, account routing/assignment, and activation - Build the account scoring and routing engine that determines account ownership (AE-owned vs. house/hedge pools), with every decision explainable and reversible. - Integrate external data sources across CRM, sales engagement, ads, and enrichment/intent vendors, and handle vendor APIs, web hooks, and rate limits. - Push signal outward, not just model it. Reverse-ETL into Salesforce, Slack digests that surface risk and opportunity before someone has to go looking, eventually activation into ad platforms. - Partner with Engineering and Product on usage data and event schemas as the product surface grows. - Design row- and column-level governance (who can see which accounts, which fields) as a first-class part of the data model, not an afterthought bolted onto Snowflake permissions later. - Debug production issues across the full path: Snowflake, dbt, orchestration, vendor APIs, web hooks, and the downstream systems (Salesforce, Slack) that depend on the output being right. - Use AI to accelerate your own build, and design the data layer (clean context, structured inputs, feedback loops) that future GTM agents (account research, pre-call prep, follow-up) will run on top of. ABOUT YOU - You’ve built data platforms, backend systems, workflow engines, or data/ML infrastructure. This is a platform-engineering role, not a dashboards-and-reporting role. - Deep experience with analytical data systems: Snowflake, dbt, and an orchestrator (Dagster, Airflow, or similar). - Comfortable writing production backend code, not just SQL. This role involves queues, workers, and stateful jobs (e.g., on AWS Batch/Lambda), not only transformation models. - Clear judgment about data modeling, schema evolution, contracts, and lineage, with an instinct for what happens to a model six months after you ship it. - You care whether a workflow actually works end to end: whether the rep trusts the Slack ping, whether the routed account was routed for a reason someone can state in one sentence. - Strong debugging instincts across a multi-system path: a broken pipeline could be a bad vendor payload, a Fivetran sync gap, a dbt model, or a downstream Salesforce write. You enjoy finding out which. - Strong understanding of how GTM teams (Sales, RevOps, Marketing) operate, enough to know why "respect the existing owner" matters more than a clever score. - Comfortable operating as the only data hire, balancing foundational architecture with urgent business needs. BONUS POINTS - You've built a customer data platform, account routing/scoring engine, or activation layer before, especially at a company with both self-serve and enterprise sales motions. - You've worked with usage-based pricing, consumption models, or PLG-plus-sales-assisted funnels. - You've supported enterprise sales motions - account scoring, territory/pool logic, pipeline analytics, or expansion reporting. - You've built or used agents for internal operations, enrichment, research, or workflow automation. - Experience with our stack, or close equivalents: Snowflake, dbt, Dagster, AWS Batch/Lambda, Fivetran, Terraform, GitHub Actions, Soda, or semantic-layer tools like Omni or Looker. - Braintrust user :) WHAT SUCCESS LOOKS LIKE - Braintrust has a governed spine (company, event, signal) that Sales, RevOps, Marketing, and Finance all build on top of. - Signals that used to get lost in a dashboard now reach a rep or a system automatically, with an audit trail for every automated decision. - Any rep can be told in one sentence why they own an account. - The platform reaches accounts that exist in the warehouse but never made it into the CRM, closing the reach gap between what you know and what you can act on. - AI is a real operating layer on top of this data: agents doing account research, prep, and follow-up on a foundation of clean, governed context, not a side experiment bolted onto a spreadsheet. - The stack stays simple enough for a startup team to run solo, while being structured enough to hand off to a business-systems team as it