Job description

SpaceX was founded under the belief that a future where humanity is out exploring the stars is fundamentally more exciting than one where we are not. Today, SpaceX is actively developing the technologies to make this possible, with the ultimate goal of making life multiplanetary and expanding consciousness to the stars. SpaceXSI is building Super Intelligence to help humanity understand the universe. Our teams are building the full Super Intelligence stack: from our Colossus supercomputers, to Grok models, to our suite of applications, including Grok Bot, Grok Build, Grok Imagine, Grok Voice, and X. We're creating a powerful flywheel where more compute trains better models, better models power more useful products, and what we learn from people using those products shapes what we build next. RESEARCH ENGINEER, SYNTHETIC DATA ABOUT THE ROLE As a Research Engineer on the Synth Data team at SpaceXAI, you’ll invent and scale the synthetic data that teaches our models new capabilities. You own the full loop: spotting a capability the model is missing, designing a data generation strategy, building reliable pipelines and verifiers, and validating that the resulting data actually moves the model. You form hypotheses about what data will help, build the systems to produce it at scale, run ablations, and read the resulting traces and training signals to understand why it worked (or didn’t). Your datasets go directly into frontier training runs, and shape the development of Grok models. WHAT YOU’LL WORK ON - Designing end-to-end synthetic data recipes that turn raw sources (traces, documents, codebases, workflows) into large, diverse, verifiable task sets. - Building strong verification and anti-hacking mechanisms so rewards are simple to compute and hard to game. - Controlling difficulty and diversity so data stays learnable and reflects real usage instead of collapsing into templates. - Running targeted ablations and analyses that demonstrate a dataset caused a capability improvement (and ruling out grader or data artifacts). - Iterating from traces and training results to close specific gaps, then turning successful one-off recipes into reusable infrastructure. - Partnering with research and other RL Data teams to translate model capability goals into concrete synthetic data plans. YOU MAY BE A FIT IF - You have research experience on synthetic or post-training data, in industry or academia. You've designed a data generation method, shipped the data into training, and can show its impact to the model you’re training. - You have strong software engineering fundamentals and have built non-trivial data pipelines, agent systems, or evaluation frameworks. - You have taste for what makes a good training task: hard enough to teach something real, simple enough to grade reliably, and resistant to reward hacking. - You are careful and systematic — understanding both aggregate metrics and close reading of raw outputs. - You enjoy turning messy, open-ended problems into repeatable, high-quality data production systems. APPLYING If there appears to be a fit, we'll reach out to schedule 2-3 short technicals. After, we'll schedule an onsite in our office, where you'll work on a small project, discuss ideas, and meet the team.

Requirements

Must-haves

Designing end-to-end synthetic data recipes that turn raw sources (traces, documents, codebases, workflows) into large, diverse, verifiable task sets.
Building strong verification and anti-hacking mechanisms so rewards are simple to compute and hard to game.
Controlling difficulty and diversity so data stays learnable and reflects real usage instead of collapsing into templates.
Running targeted ablations and analyses that demonstrate a dataset caused a capability improvement (and ruling out grader or data artifacts).
Iterating from traces and training results to close specific gaps, then turning successful one-off recipes into reusable infrastructure.
Partnering with research and other RL Data teams to translate model capability goals into concrete synthetic data plans.
You have research experience on synthetic or post-training data, in industry or academia. You've designed a data generation method, shipped the data into training, and can show its impact to the model you’re training.
You have strong software engineering fundamentals and have built non-trivial data pipelines, agent systems, or evaluation frameworks.
You have taste for what makes a good training task: hard enough to teach something real, simple enough to grade reliably, and resistant to reward hacking.
You are careful and systematic — understanding both aggregate metrics and close reading of raw outputs.
You enjoy turning messy, open-ended problems into repeatable, high-quality data production systems.
If there appears to be a fit, we'll reach out to schedule 2-3 short technicals. After, we'll schedule an onsite in our office, where you'll work on a small project, discuss ideas, and meet the team.

Nice-to-haves

None flagged.

Application prep

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

Keywords to mirror

Platform: ashby
Effort: quick