Priya Iyer

Raizen developer profile — @priya_iyer

Senior Applied Scientist at OpenAI

I build ranking and preference models with a measurement mindset: inter-rater agreement first, then the model.

Location: Seattle, WA

Ranked #2 on the Raizen verified-developer leaderboard.

Website: https://github.com/priyaiyer

Experience

• Senior Applied Scientist at OpenAI

  2022-01-10 – Present

  Preference-model training and rater-agreement tooling for assistant evals.

• Applied Scientist at Amazon

  2020-07-06 – 2021-12-20

  Search ranking; offline eval that actually moved online NDCG.

Education

• Stanford University

  M.S., Computer Science

  2018-09-01 – 2020-06-15

Verified skills

• nlp
• python
• pytorch
• ranking
• tensorflow

Projects

• Agreekit — rater agreement workbench

  Cohen/Krippendorff dashboards plus a gold-set builder. Turns noisy labels into a dataset you can retrain against.

• Pairrank — Bradley-Terry for hiring evidence

  Learns a latent quality score from pairwise reviews of candidate work. PyTorch, calibrated, with bootstrap intervals.