Job description

About Snorkel Snorkel AI is the frontier AI data lab, helping teams build the data and environments behind high-performing frontier and agentic AI. We combine technology with research-driven AI data development to create datasets, benchmarks, evals, and custom solutions for real-world AI systems. Founded out of the Stanford AI Lab in 2019, Snorkel works with leading AI labs and enterprises to move from better data to better outcomes. Excited to help us redefine how AI is built? Apply to be the newest Snorkeler! In September 2026 we raised a $350 million Series E at a $3.5 billion valuation , and we are scaling our engineering and research teams to meet demand. The role Frontier AI data is expensive to make and hard to measure. Every task we deliver is tested against the strongest models, often through many long-running agent rollouts. Your job is to make that process faster, cheaper, and more rigorous with ML and AI You will be one of the early members of ML & Research Engineering at Snorkel. You will study how frontier-grade data is generated and evaluated, form hypotheses, validate them against real production data, and ship the winners at scale. You will shape the discipline's direction, its standards, and the team that grows around it. What you'll work on Efficient agentic evals. Cut the cost of long-horizon agent evaluation with adaptive sampling, statistically grounded early stopping, model cascades, caching, and cheap-first gating. AI model routing. Route every eval and judge call to the cheapest model that clears the quality bar, with fallback, monitoring, and cost attribution. Fine-tuned small models. Fine-tune and serve open-weight models (LoRA and other parameter-efficient methods) where they match frontier quality, and know when they don't. Predictive difficulty. Build models that estimate how hard a task is for frontier systems before running a single rollout. Measurement for AI data. Build golden datasets, quantify the accuracy and calibration of LLM-as-judge systems, and make quality reproducible across projects. Research to production. Turn research prototypes into reusable, configurable components that forward deployed engineers and researchers use on every project. What you'll bring 5+ years building production ML or software systems, with end-to-end ownership from prototype to production Hands-on experience running LLM or ML workloads in production, and comfort reasoning about non-deterministic systems Deep grounding in statistics and experimentation: experiment design, hypothesis testing, sampling, and confidence intervals Strong Python and software engineering fundamentals, including testing, code review, and system design Experience designing evaluations and interpreting results rigorously A habit of finding high-impact problems before they are assigned, and clear communication with researchers, engineers, and business partners Nice to have Fine-tuning and serving open-weight models, and judging when a smaller model meets the quality bar Building LLM evaluation or experimentation platforms, model gateways, or routing systems Experience with agentic workloads, benchmarks, or RL environments A record of taking research into production: publications, open-source work, or shipped research-driven features MS or PhD in Computer Science, Machine Learning, Statistics, or a related field Why join now Frontier problems. Measure and shape the tasks designed to challenge the strongest models in the world. All AI, no plumbing. Every problem on this team is an open ML or LLM problem. Founding impact. Help define what ML engineering means at Snorkel and grow the team that carries it forward. Visible results. Your work shows up directly in the speed, quality, and cost of the data frontier AI is built on. Actual compensation will be determined based on factors including skills, qualifications, experience, and geographic location. Salary range(s) for this role $208,000 — $315,000 USD Be Your Best at Snorkel Joining Snorkel AI means becoming part of a company that has market proven solutions, robust funding, and is scaling rapidly—offering a unique combination of stability and the excitement of high growth. As a member of our team, you’ll have meaningful opportunities to shape priorities and initiatives, influence key strategic decisions, and directly impact our ongoing success. Whether you’re looking to deepen your technical expertise, explore leadership opportunities, or learn new skills across multiple functions, you’re fully supported in building your career in an environment designed for growth, learning, and shared success. Snorkel AI is proud to be an Equal Employment Opportunity employer and is committed to building a team that represents a variety of backgrounds, perspectives, and skills. Snorkel AI embraces diversity and provides equal employment opportunities to all employees and applicants for employment. Snorkel AI prohibits discrimination and harassment of any type on the basis of race, color, religion, age, sex, national origin, disability status, genetics, protected veteran status, sexual orientation, gender identity or expression, or any other characteristic protected by federal, state, or local law. All employment is decided on the basis of qualifications, performance, merit, and business need. We will ensure that individuals with disabilities are provided reasonable accommodation to participate in the job application or interview process, to perform essential job functions, and to receive other benefits and privileges of employment. Please contact us to request accommodation.

Requirements

Must-haves

What you'll work on
Efficient agentic evals. Cut the cost of long-horizon agent evaluation with adaptive sampling, statistically grounded early stopping, model cascades, caching, and cheap-first gating.
AI model routing. Route every eval and judge call to the cheapest model that clears the quality bar, with fallback, monitoring, and cost attribution.
Fine-tuned small models. Fine-tune and serve open-weight models (LoRA and other parameter-efficient methods) where they match frontier quality, and know when they don't.
Predictive difficulty. Build models that estimate how hard a task is for frontier systems before running a single rollout.
Measurement for AI data. Build golden datasets, quantify the accuracy and calibration of LLM-as-judge systems, and make quality reproducible across projects.
Research to production. Turn research prototypes into reusable, configurable components that forward deployed engineers and researchers use on every project.
5+ years building production ML or software systems, with end-to-end ownership from prototype to production
Hands-on experience running LLM or ML workloads in production, and comfort reasoning about non-deterministic systems
Deep grounding in statistics and experimentation: experiment design, hypothesis testing, sampling, and confidence intervals
Strong Python and software engineering fundamentals, including testing, code review, and system design
Experience designing evaluations and interpreting results rigorously
A habit of finding high-impact problems before they are assigned, and clear communication with researchers, engineers, and business partners

Nice-to-haves

Fine-tuning and serving open-weight models, and judging when a smaller model meets the quality bar
Building LLM evaluation or experimentation platforms, model gateways, or routing systems
Experience with agentic workloads, benchmarks, or RL environments
A record of taking research into production: publications, open-source work, or shipped research-driven features
MS or PhD in Computer Science, Machine Learning, Statistics, or a related field
Why join now
Frontier problems. Measure and shape the tasks designed to challenge the strongest models in the world.
All AI, no plumbing. Every problem on this team is an open ML or LLM problem.
Founding impact. Help define what ML engineering means at Snorkel and grow the team that carries it forward.
Visible results. Your work shows up directly in the speed, quality, and cost of the data frontier AI is built on.

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

aimlllmpythonmachine learning

Platform: greenhouse
Effort: quick