Loading...
Mid LevelFull-timeRustAWSAIML
Salary Range
$102,000 - $138,000 /year
This salary is estimated based on similar roles. The actual salary may vary.
About the Role
We are looking for a Software Engineer to join the Human Alignment Team within Future of Computing Research to build the infrastructure, data systems, and evaluation foundations for next-generation multimodal models.
Key Responsibilities
- →Build evaluation and data foundations for next-generation personalized and multimodal AI systems
- →Partner closely with researchers to turn fuzzy behavioral questions into rigorous evals, datasets, rubrics, and scorecards
- →Design and implement human-data pipelines, grader systems, and experiment infrastructure for product-grounded research
- →Create evaluation frameworks for subjective, contextual, and long-horizon behaviors
- →Develop reproducible pipelines for collecting, processing, joining, and analyzing multimodal signals from real-world studies and product usage
- →Help define what should count as meaningful progress, and build the systems that let the team measure it with confidence
- →Work across research, safety, design, and engineering to ensure that what we optimize for is both technically sound and human-centered
- →Prototype quickly, iterate on measurement frameworks, and improve the team’s ability to debug, compare, and trust behavioral results
- →Shape the infrastructure and methodology that future OpenAI products will rely on for personalization, adaptation, and evaluation
Requirements
- ✓Have strong software engineering fundamentals and experience building data, backend, ML, or evaluation systems
- ✓Have excellent research taste and strong judgment about what is worth measuring, how to measure it, and when a metric is misleading
- ✓Enjoy working on ambiguous, early-stage problems where the hardest part is often defining the right evaluation rather than implementing the obvious one
- ✓Have experience with human-in-the-loop systems, annotation pipelines, experimentation platforms, or evaluation tooling
- ✓Are rigorous about data quality, reproducibility, metric design, and empirical correctness
- ✓Are motivated by human-centered AI and excited by the challenge of measuring behaviors that are subtle, contextual, and difficult to benchmark
- ✓Want to help define the research infrastructure behind a new category of AI products
Free to sign up. Takes 30 seconds.
