- Location
- United States
- Workplace
- Remote
- Employment
- Full time
- Level
- Mid level
- Posted
- 3 months ago
About this role
Company Overview:
We are building Protege to solve the biggest unmet need in AI - getting access to the right training data. The process today is time intensive, incredibly expensive, and often ends in failure. The Protege platform facilitates the secure, efficient, and privacy-centric exchange of AI training data.
Solving AI’s data problem is a generational opportunity. We’re backed by world-class investors and already powering partnerships with some of the most ambitious teams in AI. The company that succeeds will be one of the largest in AI - and in tech.
We’re a lean, fast-moving, high-trust team of builders who are obsessed with velocity and impact. Our culture is built for people who thrive on ambiguity, own outcomes, and want to shape the future of data and AI.
About DataLab
DataLab exists because truly useful data is rare - and the frontier of AI development only moves forward when high-quality data makes it possible.
We believe data is one of the most underdeveloped layers of the AI stack. Our work focuses on building and evaluating high-value datasets grounded in real-world workflows and economically meaningful tasks.
We work across multiple domains to create safe, high-fidelity datasets that preserve the structure and context needed to train advanced AI systems.
Our research spans data quality, evaluation design, privacy-preserving transformation, workflow reconstruction, and task-grounded AI training data.
At DataLab, applied research is tightly connected to real-world deployment. Researchers work directly with large-scale datasets, production systems, and frontier AI training problems.
Role Overview
Data is the foundation of AI performance, and we believe model quality starts with data quality. As AI systems become more agentic, a critical challenge is understanding which real-world datasets, tasks, and environments actually lead to better model behavior.
We’re seeking a Machine Learning Researcher focused on RL and agentic systems to help define, design, and evaluate the datasets, tasks, environments, and benchmarks used to assess advanced AI systems. In this role, you’ll work closely with research and engineering teams to translate real-world workflows into high-value datasets and evaluation assets: structured tasks, interactive environments, benchmark suites, and quality scorecards that help us understand how models perform in realistic settings.
You’ll help define what “high-quality agentic data” means in practice, using statistical, computational, and ML-driven methods to evaluate dataset quality, task design, environment fidelity, and downstream model performance. You’ll work on the core problems of benchmarking real-world data, measuring how well models perform on that data, and designing RL-style or agentic environments that capture the structure of meaningful work.
This is an ideal role for someone with a strong machine learning background who is excited by reinforcement learning, agentic systems, evaluation, and the role of data in shaping model behavior. You should be excited by the opportunity to build the datasets and benchmarks that help define what high-quality real-world data looks like for frontier AI systems.
What You’ll Do
Design and build datasets, tasks, and environments
Design and build datasets, tasks, environments, and evaluation assets for benchmarking agentic systems and multi-step model behavior.
Translate real-world workflows into structured tasks, interaction traces, trajectories, stateful environments, and verifiable outcomes that can be used to evaluate advanced AI systems.
Develop frameworks for evaluating real-world data quality
Develop frameworks that assess diversity, realism, coverage, fidelity, informativeness, and downstream usefulness of datasets for agentic systems.
Build quality scorecards and evaluation methods that make dataset strengths, weaknesses, and failure modes legible across teams.
Benchmark model behavior in RL and agentic settings
Evaluate planning, tool use, robustness, recovery from failure, task completion, and generalization behavior in RL-style or agentic environments.
Connect model failures back to concrete dataset, environment, or task-design gaps and recommend improvements grounded in empirical evidence.
Build scalable evaluation and validation tooling
Contribute to tools and systems that automate dataset validation, environment generation, rollout analysis, benchmark construction, and evaluation workflows.
Improve internal infrastructure for reproducible experimentation, benchmark management, and evaluation quality.
Partner across research, engineering, and product
Collaborate closely with research and engineering teams to identify data bottlenecks, improve evaluation methodology, and shape internal best practices around task-grounded AI training data.
Represent DataLab’s perspective in cross-functional discussions around dataset quality, benchmark design, and frontier agentic-system evaluation.
What Success Looks Like
Near-term: establish a strong evaluation baseline
Create clear benchmark frameworks, evaluation assets, and dataset-quality scorecards that help Protege reason about how real-world data impacts advanced agentic systems.
Use rigorous evaluation methods to identify meaningful dataset improvements, improve benchmark fidelity, and sharpen the company’s understanding of what high-impact agentic data actually looks like in practice.
What You Bring
PhD or equivalent Master’s Degree + 4+ years industry experience in machine learning, computer science, statistics, engineering, mathematics, economics, or related quantitative fields.
Strong understanding of AI model training pipelines, evaluation methodology, and the role of data in shaping model performance.
Experience working with large, unstructured, or semi-structured datasets used to train or evaluate ML systems.
Experience with reinforcement learning, sequential decision-making, agentic systems, tool-using models, or multi-step model evaluation.
Experience designing tasks, benchmarks, environments, simulations, or evaluation frameworks for real-world model behavior.
Strong intuition for realism, coverage, difficulty, fidelity, and meaningful outcome structure in datasets.
Strong experimental design, evaluation, benchmarking, and data-validation skills.
High ownership and ability to independently identify and solve high-impact problems.
Nice to have
Experience developing evaluation frameworks or performance metrics for datasets, agentic systems, or training data.
Experience translating real-world workflows into structured tasks or environments for model evaluation.
Experience with RLHF, RLAIF, imitation learning, reward modeling, online or offline RL, or related methods.
Experience with Harbor or other agent evaluation frameworks.
Publications or open-source contributions in reinforcement learning, agents, evaluation, or data-centric AI.
Experience collaborating cross-functionally with product, infrastructure, or partnership teams.
Experience with synthetic data generation, trajectory generation, or simulation-based environments.
Protege's Values
Pass the Loved Ones' Test
We act with integrity and do the right thing - especially when it's hard and no one is watching.
Always Find a Way
We are resourceful, resilient builders who solve hard problems and push through obstacles.
Go Fast and Grow Fast
Velocity matters. We move with urgency, learn quickly, and continuously improve as individuals and as a company.
Practice Kindness and Candor
We communicate directly and respectfully, building trust through honest feedback and genuine care for one another.
Deliver Together
We win as one team. Collaboration, accountability, and shared ownership drive our success.
Own the Outcome. Hone the Craft.
We take pride in our work, sweat the details, and continuously raise the bar for excellence.
As published by Protege. Applications are handled on their site.
Skills this posting mentions
About Protege
The biggest unmet need in AI today is getting access to the right training data. Data holders often don’t know where to start and are rightly concerned about governance, intellectual property, and security implications. AI companies can spend years finding and negotiating access to the data they need. Protege is solving these problems by providing an easy-to-use platform to connect data holders with vetted data users.
All 16 openings at ProtegeOne click, then it is written
Apply to Protege with a resume written for this role.
Queue Machine Learning Researcher, RL & Agentic Systems and I read the posting, rewrite your resume against it, draft the cover letter, and score the fit. Then you press send, or press one button and I fill in Protege’s form for you.
01Drop your resume
A PDF or a LinkedIn URL. About a minute, once.
02I rank the openings
Every weekday morning, the live catalog scored against your history. Up to 20 worth your time, not two hundred links.
03Each one is written up
Resume rewritten for the posting, a cover letter, a fit score. Press send, or let me fill in the form.
- 25 sent a week, free
- No card
- Nothing sent until you say so
More roles at Protege
See all- 5 weeks ago
- 6 weeks ago
- 6 weeks ago
- 6 weeks ago
- 6 weeks ago
- 2 months ago
Similar roles elsewhere
See more- Today
Machine Learning Engineer, Relevance and Personalization
AirbnbUnited States
$166k - $195k/yrData and ML - Today
Senior Machine Learning Engineer, Ads Optimization
RedditRemote - United StatesRemote
$217k - $303k/yrSeniorData and ML - Today
Sr. Staff Data Scientist - Ads Measurement, Signals, Privacy
RedditRemote - United StatesRemote
$233k - $326k/yrStaffData and ML - Today
Staff Data Scientist - Ads Measurement, Signals, Privacy
RedditRemote - United StatesRemote
$217k - $304k/yrStaffData and ML - Today
Senior Staff Data Scientist - Consumer Experimentation
RedditRemote - United StatesRemote
$233k - $326k/yrStaffData and ML
Put this to work
Paste your career in once. Every application after that is written for you.
Drop a resume or a LinkedIn URL. I rank the live openings against it, rewrite the resume and write a cover letter for the best of them, and fill in the employer's form when you press the button. You read, you decide what goes out.
01Drop your resume
A PDF or a LinkedIn URL. About a minute, once.
02I rank the openings
Every weekday morning, the live catalog scored against your history. Up to 20 worth your time, not two hundred links.
03Each one is written up
Resume rewritten for the posting, a cover letter, a fit score. Press send, or let me fill in the form.
- New matches ranked and written before you are up.
- Every bullet stays inside what your history supports. Nothing invented.
- Queued, submitted, interviewing, offer: one screen, not a spreadsheet.
500 free credits on sign-up. No card. Nothing is sent until you say so.
Listed from the job board Protege publishes. Refolk is not the employer and does not handle their hiring. Applications go to Protege directly.