Own a core surface
from line one.
Every AI team hits the same wall: the model is ready, the data isn't. Generation and LLM-as-judge finally cost cents per thousand rows, and open table formats make "connect to the warehouse instead of copying it" actually realistic. Datastrap sits on that shift — a thin, zero-copy control plane that expands a tiny seed into a clean training set inside the customer's own warehouse, with a judge doing the boring 80% and humans only on the hard rows. We're building the orchestration layer for the data flywheel before the warehouses bolt on a worse version of it. If you want to define an infra category early, this is the moment.
See open roles →Founding Engineer — ML / LLM Systems
COREMission — Own the generation + judge loop, the trust boundary the entire product rests on. Make the judge calibrated, diverse, and auditable at scale.
Must-haves: Strong Python; hands-on with LLM APIs, structured outputs, embeddings; intuition for eval/quality loops; comfort owning a system end-to-end with no playbook.
Why it matters: If the judge is wrong, we silently poison customer models. This role is the product's credibility.
Founding Full-Stack Engineer
COREMission — Build the zero-copy connector and the orchestration spine — read seed, fan out generation, write approved rows back safely, never persist payload.
Must-haves: TypeScript/Next.js + Python/FastAPI; warehouse/Postgres internals; async job systems; a healthy fear of writing into a customer's production table.
Why it matters: Write-back is the most dangerous operation we perform. You make it idempotent, staged, and boringly reliable — that reliability turns a demo into a contract.
Product Designer / DevRel
Mission — Own the "magic moment" — seed to first synthetic row in under 3 minutes — and the keyboard-only review UI, then take the story to the communities our buyers live in.
Must-haves: Strong product/interaction design for technical users; ships in-code or hands off pixel-tight; writes well enough to demo in MLOps and LLM-dev communities.
Why it matters: Our moat is opinionated UX plus a quality loop people trust. You shape both.
Founding GTM / Growth
Mission — Turn demand-proof into a pipeline — find funded AI teams already paying for labeling/evals, run the priced experiments, convert design partners into paying accounts.
Must-haves: Sold or marketed dev/infra tools to technical buyers; comfortable in Slack/Discord/Reddit and cold outbound; data-driven on funnel and CAC; credible on evals and fine-tuning.
Why it matters: A great product with no distribution is a hobby. You build the repeatable motion.
ML Engineer — Data Quality & Audit
Mission — Own the continuous sampled audit, diversity scoring, and threshold auto-tuning that keep judge precision honest as usage grows.
Must-haves: ML/data-eng background; comfortable with embeddings, dedup, calibration, statistical sampling; rigorous about measurement over vibes.
Why it matters: The audit loop is the early-warning system we sell to enterprise buyers. This role keeps the number on the slide real.
Small, fast, and honest about it.
We're pre-seed: a handful of people, no layers, and a direct line to the founders — you'll ship to real design partners in your first weeks. We write our disagreements down instead of smoothing them over, default to the cheapest tool that clears the quality bar, and treat the judge's trustworthiness as non-negotiable. Compensation is equity-heavy with a livable cash floor — founding-team ownership, not BigCo salary. If you need a big team and a clear playbook, this isn't it yet. If you want to build the playbook, come talk to us.
How to apply
Tell us one thing you've shipped that you're proud of, and one thing about Datastrap you'd do differently. No cover letters.