Drimva

Contra Labs

Research Scientist, Human Data & Evaluation

$150K–$200K

San Francisco · Remote · Full-time · Posted Aug 17, 2026

Experience
3–5+ years of relevant research experience.
Education
Master’s, PhD, or equivalent industry research depth in HCI, Human-Centered AI, ML, design research, or a related field.

About the job

Independently lead human-centered AI research projects, from defining research questions and methodology to analyzing results and delivering recommendations, evaluations, benchmarks, and reports.

What you'll do

  • Develop research proposals, study designs, human-data collection strategies, and evaluation protocols.
  • Analyze qualitative and quantitative data and turn findings into insights, recommendations, and research outputs.
  • Own research engagements from scoping through delivery, collaborating with internal teams and clients.

What they ask for

  • 3–5+ years of relevant research experience, with a Master’s, PhD, or equivalent industry research depth in HCI, Human-Centered AI, ML, design research, or a related field.
  • Strong qualitative and quantitative research skills, including experience with human evaluation, experimentation, data collection, UX/HCI, annotation, or model behavior research.
  • Ability to independently design and execute rigorous research and turn findings into actionable recommendations.
  • Comfort analyzing data with Python and/or SQL; based in San Francisco, California.

Nice to have

  • Familiarity with modern ML and generative AI systems; understanding of RLHF, post-training, preference data, or model evaluation.
  • Experience with multimodal or creative systems, prompt research, large-scale human-data programs, AI research teams, original research, or startup work.

Pay

$150,000 - $200,000 USD base salary

Tools

Python, SQL

AI focus

Research and evaluation for creative and multimodal AI, including human preference data, benchmarking, post-training, and model behavior.

Model-written summary · the original posting is authoritative

First seen Oct 27, 2026 · Last verified Nov 5, 2026 (UTC)