Zingle AI Labs is an applied research lab for data agents, based in San Francisco. We're backed by Accel and Nexus, along with people from Anthropic, Scale AI, Cursor, MinIO, Hasura, and Postman.
Reinforcement learning usually breaks down at the reward function. In most domains, deciding what "correct" means takes rubrics, human raters, and months of disagreement. Data is different. Correct means the numbers match: row counts, sums, null profiles, distributions. The reward is verifiable, binary, and already sitting in the warehouse.
Data's hard problem was never the reward. It was the cost of an attempt. Checking a single change at production scale costs as much as the job itself, so agents got one try and a human did the rest.
We made attempts cheap. Zingle gives agents a faithful scale model of a real warehouse (whole entities, full history, joins intact, per-entity numbers exact) and enough compute to run 100+ attempts an hour, each one checked against ground truth. Teams bring their own agents. We provide the environment and the verdict. Next, we're extending the same environment to train models for data work.
We're currently live with 4 enterprises across US (US's top Fintech and sports-tech data teams are building their own agents on top of us)
About the role:
You'll be our first product hire, working directly with the founders and our research and engineering team. You'll define how RL environments become everyday infrastructure for data teams.
Spend time with data leaders and the engineers running agents in production, and find exactly where agents still guess.
Own the roadmap across the environment, the sandbox, compute, and the verdict layer, and decide what we build next and why.
Turn research capabilities into products that data teams trust with real warehouses.
Write clear specs, ship with engineering, and measure outcomes in live deployments.
You may be a good fit if you:
Have a strong technical background. This is non-negotiable.
Have 8+ years of experience in data engineering, data platforms, developer tools, or product.
Understand the modern data stack deeply: warehouses, dbt, Spark, orchestration.
Use AI coding agents like Claude Code, Codex, or Cursor, and have opinions about where they fail.
Can hold a room with a CDO and a whiteboard session with an engineer on the same day.
Write clearly and think in systems.
Strong candidates may also have:
Exposure to evals, RL, or verifiable-reward setups
Experience with warehouse migrations or compute optimization
Early-stage startup experience
Compensation and logistics:
$300K - $350K Base + 0.6% - 0.8% equity
Direct work with data and AI leaders at Fortune 500 companies
Remote, with regular travel to San Francisco
Location: Remote, US (NYC or Chicago preferred)