Snorkel AI · , New York City, NY (Hybrid) · San Francisco, CA (Hybrid)
$220k–$300k a year · Hybrid · Lead & above
Medium competition (est.)
Filed H-1B · FY25Filed 4 H-1B LCAs in FY2025
Posted 2 months ago
Working fluency in modern LLM post-training: SFT, preference data (RLHF/RLAIF), RLVR, reward modeling, and how data composition affects model capability. Must be able to hold a substantive conversation with a research scientist without an interpreter
Familiarity with agentic systems and the current agentic eval landscape (e.g. SWE-bench-style coding evals, terminal/computer-use benchmarks, tool-use and long-horizon task evaluation) and an informed view on where they fall short
Track record building product frameworks from zero — prioritization models, roadmap artifacts, intake processes — in an environment with no existing playbook
Experience operating across research, GTM, and operations simultaneously, with evidence of driving decisions through influence rather than authority
Direct customer-facing experience with highly technical buyers; ability to run a discovery conversation with an ML researcher or post-training lead and convert it into a roadmap commitment
Quantitative rigor: can size a market, model unit economics of a data program (cost per trajectory/task/environment), and defend prioritization with numbers
Strong technical foundation — comfortable reading research papers, discussing training dynamics with scientists, and reasoning about data pipelines end-to-end.
Ability to write clearly for two audiences at once — internal research/ops and external frontier lab stakeholders
Excellent analytical instincts — able to define success metrics for products that live close to research, where outcomes are often indirect.
Preferred Qualifications
Prior experience at an AI data/environments company or inside a frontier lab's data, post-training, or evals org
Has built or specified RL environments — sandboxed/containerized task environments, verifiable reward design, task generation, environment scaling and reproducibility
Direct experience with agentic coding or computer-use data specifically: trajectory collection, rubric design, verifier construction, failure-mode taxonomy
Hands-on technical ability — can write Python, query data, run a model, and prototype an eval without engineering support
Experience selling or delivering into frontier labs, with an existing network among post-training, evals, or data acquisition leads
Experience structuring academic or research partnerships, including co-development of datasets or benchmarks
Published research, open-source datasets/benchmarks, or public writing that establishes credibility with the research community
Experience with pricing and packaging for bespoke or semi-standardized data contracts, and the tension between custom deals and repeatable product
Competitive intelligence muscle — has run structured win/loss or market mapping in a fast-moving, opaque market
Prior founding-PM or 0→1 experience at a company between Series B and IPO
Domain depth in one or more target verticals for agentic data (software engineering, enterprise workflows/CRM-ERP automation, finance, healthcare)
Experience with human-in-the-loop data pipelines, annotation quality systems, or synthetic data generation at scale.
Track record of leading large, cross-team initiatives without formal authority.
Actual compensation will be determined based on factors including skills, qualifications, experience, and geographic location.
Sponsorship data shows employers' past filings in public US government records (DOL, USCIS, E-Verify) and is not a promise that any job will be sponsored. Immigration rules change. Verify your situation with your DSO or an immigration attorney.