Research / Data Engineer – AI & Reinforcement Learning

CGS; San Francisco (CA); In loco; USD 150.000 - 210.000; Tempo pieno

Descrizione del lavoro

CGS is seeking a Research / Data Engineer to scale our post‑training data and RL environments for frontier AI labs. You will bridge research and engineering, taking ownership from infrastructure provisioning to dataset curation and evaluation.

You will develop and experiment with ML techniques, build scalable data pipelines, and contribute to high‑quality training datasets used to push next‑gen AI models. This role is hands‑on and research‑driven.

Competenze

• 3–7 years of ML research, ML engineering, or a closely related field.
• Demonstrated research expertise in at least one area such as model post‑training, reinforcement learning, model evaluation, or interpretability.
• Strong engineering skills alongside research experience; excellence in only one area is not sufficient.
• Practical experience post‑training models and curating high‑quality post‑training datasets.
• Strong Python and PyTorch skills, with the ability to understand unfamiliar codebases quickly.
• Experience developing, debugging, and optimising machine learning pipelines.
• Strong judgement regarding data quality and task design, or the ability to develop this expertise quickly.
• Genuine, self‑driven interest in AI research demonstrated through publications, personal projects, open‑source contributions, or similar work.

Mansioni

• Own post‑training workflows: end‑to‑end process from infrastructure provisioning to dataset curation and QA.
• Build reinforcement learning environments and long‑horizon tasks across multiple domains.
• Develop high‑quality training data that addresses the needs of leading AI research laboratories.
• Scale data generation with self‑improving developer processes and automated workflows.
• Drive research and experimentation to improve post‑training methods and data quality.
• Maintain high data standards to ensure useful tasks for advancing AI model capabilities.
• Build robust ML pipelines to support model training, experimentation, and data generation.

Conoscenze

Python

PyTorch

ML engineering

Data pipelines

Descrizione del lavoro

Research / Data Engineer – AI & Reinforcement Learning

Working hours: Full-time, approximately 9:00 AM–9:00 PM, 6 days per week

Visa: US citizens, green card holders, F-1, H-1B, O-1, and TN candidates considered

About the Company

Our client is a cash-flow-positive, seed-stage AI company building high-quality, long-horizon reinforcement learning (RL) environments and post-training datasets for frontier AI laboratories.

The company actively works with leading US AI labs and top international players, helping advance the capabilities of next-generation AI models.

While traditional data providers often approach post-training data as a matter of increasing headcount, this team takes a research-driven approach. Its goal is to scale data production through self-improving processes, enabling increasingly sophisticated training environments without relying on linear increases in staffing.

You'll join a small, exceptionally strong team of researchers and engineers working at the intersection of machine learning research, data generation, and AI development. The company recently released a benchmark designed to measure how effectively models can iterate within an ML research loop.

About the Role

The demand for the company’s post-training data exceeds its current production capacity. As a Research / Data Engineer, you’ll help scale the core team by combining hands‑on machine learning research with high-quality data production.

This role brings together two disciplines that are often separated: research and engineering. You’ll not only develop and experiment with ML techniques but also build the infrastructure, environments, and datasets required to improve frontier AI models.

The ideal candidate has a genuine research specialisation in areas such as reinforcement learning, model post‑training, evaluation, or interpretability, alongside strong engineering capabilities. You’ll take ownership of the process end to end, from provisioning infrastructure to designing realistic tasks and curating high-quality training data.

Key Responsibilities

• Own post-training workflows: Manage the end‑to‑end process, from infrastructure provisioning and model post‑training to dataset curation and quality assurance.
• Build reinforcement learning environments: Design and develop sophisticated RL environments and long‑horizon tasks across multiple domains, including scientific and ML research workflows.
• Develop high‑quality training data: Create realistic, challenging, and valuable datasets that address the needs of leading AI research laboratories.
• Scale data generation: Build self‑improving developer processes and automated workflows that increase data production without simply adding more people.
• Drive research and experimentation: Apply machine learning research to improve post‑training methods, evaluation processes, and the quality of generated data.
• Maintain high data standards: Identify which tasks are realistic, economically valuable, and genuinely useful for advancing AI model capabilities.
• Build robust ML pipelines: Develop and optimise the engineering systems that support model training, experimentation, and data generation.

Requirements

• Approximately 3–7 years of relevant experience in machine learning research, ML engineering, or a closely related field.
• Demonstrated research expertise in at least one area such as model post‑training, reinforcement learning, model evaluation, or interpretability.
• Strong engineering skills alongside research experience. Candidates must demonstrate both; excellence in only one area is not sufficient.
• Practical experience post‑training models and curating high‑quality post‑training datasets.
• Strong Python and PyTorch skills, with the ability to understand unfamiliar codebases quickly and solve challenging technical problems.
• Experience developing, debugging, and optimising machine learning pipelines.
• Strong judgement regarding data quality and task design, or the research ability to develop this expertise quickly.
• Genuine, self‑driven interest in AI research, demonstrated through publications, personal projects, open‑source contributions, technical writing, or similar work.

Ideal Background

We’re looking for candidates who combine research depth with the ability to build and ship practical systems.

Particularly strong candidates may have:

• Published research or made meaningful contributions to reinforcement learning, model post‑training, interpretability, or AI evaluation.
• Built and released notable side projects, open‑source tools, or ML infrastructure.
• Developed high‑performance libraries, influential benchmarks, or research papers that have attracted significant citations.
• Combined ML expertise with domain knowledge in areas such as neuroscience, physics, chemistry, or computer systems.
• Demonstrated the ability to translate research ideas into functioning tools, datasets, or production‑ready workflows.
• A strong record of independent experimentation and rapid technical development.

The strongest candidates are curious, research‑driven, and highly motivated to push the boundaries of what AI systems can learn.

This role is not suited to a purely data‑engineering or data‑services profile. Candidates who are strong engineers but lack meaningful ML research experience, or strong researchers without the engineering ability to implement their ideas, are unlikely to be a fit. The team values rapid learning, technical depth, and a demonstrated ability to produce exceptional work.

Interview Process

• Introductory call with a founder.
• Technical interview focused on independent problem‑solving, including a machine learning exercise using PyTorch.
• Practical work trial.
• Offer.

Why Join?

• Work at the frontier of AI: Contribute to post‑training and reinforcement learning systems used by leading AI research laboratories.
• Combine research and engineering: Take ownership of both the scientific questions and the systems needed to answer them.
• Solve fundamental scaling challenges: Develop self‑improving approaches to data generation rather than relying on increasing headcount.
• Work with an exceptional team: Collaborate with a small group of highly capable researchers and engineers.
• Make a measurable impact: Help shape the environments, datasets, and workflows that influence the capabilities of next‑generation AI models.

If you’re an engineer with genuine ML research depth who wants to turn ambitious research ideas into scalable, high‑quality training systems, this is an opportunity to work on some of the most challenging problems in frontier AI.