Research Engineer

Neon Blue•San Francisco, CA
•$170,000 - $250,000•Onsite

About The Position

Neon Blue is a seed-stage company in San Francisco that builds simulated audiences grounded in real human studies, helping teams test product, pricing, and marketing decisions before spending money. The Research Engineer is a senior individual-contributor role on a small research team, reporting to the CTO. The role combines experimental research with building the data and evaluation systems that support it, with high ownership of work from hypothesis through results that can be used in the product and shared publicly. The Research Engineer will develop and test methods for building and controlling simulated people, including approaches beyond prompting; ground those methods in human studies; and build reproducible evaluations comparing simulated responses with real human responses. The role owns datasets, evaluation harnesses, pipelines, versioning, and provenance, and includes investigating failure modes, uncertainty, and methodological limits. The engineer will collaborate with the research team to publish research and take useful results into production. In the first 90 days, the engineer will learn the current research stack, data, and evaluations; work with the CTO to define what improvement means for upcoming experiments; run a focused, reproducible experiment with a clear baseline; own the evaluation harness comparing simulated and real answers; and document methods and follow-up experiments. This is a full-time, in-person role in San Francisco and is not a manager position.

Requirements

  • Solid Python skills.
  • Hands-on experience training, fine-tuning, or controlling language models and running experiments on GPUs.
  • Experience building data pipelines, evaluation harnesses, or research infrastructure.
  • Care about data quality and reproducibility.
  • Experimental rigor, including falsifiable hypotheses, meaningful baselines, and evaluations that are difficult to game.
  • Ability to read papers, reproduce methods, explain their assumptions, and improve on them.
  • Interest in research that reaches a product and public work.
  • Comfort working end to end in a small, in-person team.

Nice To Haves

  • Published research, public benchmarks, or open evaluation infrastructure.
  • Work controlling simulated people beyond prompting, or work on persona and behavioral consistency.
  • Background in computational social science, psychometrics, or behavioral modeling.
  • Experience taking a research system into monitored production.

Responsibilities

  • Design and run experiments on methods for building and controlling simulated people, including methods beyond prompting.
  • Ground methods in real human studies and build evaluations comparing simulated answers with real people's answers.
  • Own datasets, evaluation harnesses, and pipelines that make comparisons a regular part of the work, including versioning and provenance.
  • Work with the research team to turn research into public work.
  • Investigate failure modes, including uncertainty and the limits of a method.
  • Take useful research results into production with the team.
  • Document methods, negative results, and next questions so the work is reproducible.
  • Learn the current research stack, data, and evaluations, and agree with the CTO on goals for upcoming experiments.
  • Run a focused experiment on ways to better control or ground simulated people, with a clear baseline and reproducible results.
  • Own the evaluation harness comparing simulated and real answers.

Benefits

  • Equity
© 2026 Teal Labs, Inc
Privacy PolicyTerms of Service