Research Engineer, Synthetic Data (Pacific Time)

Pragmatike; United States; Hybrid; USD 120,000 - 190,000; Full time

Job summary

Pragmatike is seeking a hands-on research engineer to build a synthetic data pipeline that turns professional workflows into useful AI training tasks. This role spans task design, generation systems, evaluation and analysis of what models learn, taking early research through to dependable pipelines.

You will collaborate with domain experts to create varied, realistic tasks, design generation approaches, and build tools to mutate and validate data, aiming for reliable operation across time zones.

Qualifications

• Proficiency with Python, Docker, and Linux.
• Experience with synthetic data research methods and an understanding of the quality and limitations of synthetic data.
• Experience building synthetic data pipelines end to end, including work without a prescribed roadmap.
• Clear written and spoken English and ability to collaborate across time zones.

Responsibilities

• Work with subject‑matter experts to develop training tasks across professional and technical domains.
• Design generation approaches that produce varied, realistic, and learnable tasks.
• Build tools to mutate, validate, and improve generated tasks.
• Evaluate model and agent results to identify what tasks teach and where performance breaks down.
• Create measures for task diversity, realism, learnability, and related qualities.
• Develop pipeline methods and systems from initial research into reliable operation.

Skills

Synthetic data research methods

End-to-end data pipelines

Collaboration across time zones

English communication

Tools

Python

Docker

Linux

Job description

Role overview

Build a synthetic data pipeline that turns professional and technical workflows into useful training tasks for AI agents. This hands‑on research engineering role spans task design, generation systems, evaluation, and analysis of what models learn; it includes taking early research through to dependable pipelines in areas where the path is not fully defined.

Responsibilities

• Work with subject‑matter experts to develop training tasks across professional and technical domains.
• Design generation approaches that produce varied, realistic, and learnable tasks.
• Build tools to mutate, validate, and improve generated tasks.
• Evaluate model and agent results to identify what tasks teach and where performance breaks down.
• Create measures for task diversity, realism, learnability, and related qualities.
• Develop pipeline methods and systems from initial research into reliable operation.

Requirements

• Proficiency with Python, Docker, and Linux.
• Experience with synthetic data research methods and an understanding of the quality and limitations of synthetic data.
• Experience building synthetic data pipelines end to end, including work without a prescribed roadmap.
• Experience with environments, evaluations, or benchmarks.Clear written and spoken English and ability to collaborate across time zones.

Nice to have

• Careful attention to subtle inconsistencies and edge cases in generated data.
• First‑principles reasoning about task design, scoring, and failure modes.
• Comfort working in early‑stage or otherwise unstructured settings.

Benefits and work setup

• Full‑time role, with on‑site options in San Francisco or Singapore, or remote work from Singapore or US Pacific time zones. Remote schedules should overlap roughly 70-80% with business hours in one of those regions; exact hours are agreed during hiring.

Technical aptitude and learning potential are emphasized over years of experience, and applicants need not meet every listed qualification.