Miju Labs

← All buyers · OpenAI

Alexandra Barr

Program Manager, Human Data (LinkedIn headline 'OpenAI | Human Data') · OpenAI

medium confidenceTier A · Talk to firstFrontier labsimage · text

Barr is a programme manager on OpenAI's Human Data team and one of the few people there who talks about the team in public — she recruits for it openly on X and LinkedIn. She shows up in the credits of the products where human data was decisive: a foundational contributor on deep research, a GDPval co-author, and in the Data & Evaluation group for ChatGPT Images. She is a realistic first point of contact into OpenAI's human-data function.

Background

Barr's public footprint is entirely OpenAI-era. She identifies herself simply as 'OpenAI | Human Data' and posts recruiting threads for the team, describing it as one of the best teams in the organisation to join and circulating its Program Manager, Human Data opening. That visibility is unusual for the function and makes her unusually approachable relative to her seniority.

Her credit pattern is the interesting part. On deep research (Feb 2025) she is listed among 'Foundational Contributors' alongside research leads, not in a support section — the human-data contribution was treated as foundational to the model. She then co-authored GDPval (Oct 2025), OpenAI's expert-graded benchmark of real occupational work, and appears in the Data & Evaluation group for ChatGPT Images / gpt-image-1.5 (Dec 2025), which is the image-generation data and eval organisation.

What they run now

Human-data campaigns that feed both training and evaluation. Concretely: she was part of the data work behind deep research, part of the GDPval author group that built and ran expert grading of occupational deliverables, and part of the Data & Evaluation group for ChatGPT Images — which means she has touched the image-generation data pipeline, not just text. She also does visible hiring for the Human Data team, which implies she owns programme capacity, not just a single campaign.

What they buy

She specifies and runs campaigns rather than signing contracts: task design, annotator qualification, rubric writing, preference collection and expert grading. Evidence puts her on text/agentic data (deep research), expert occupational deliverables and grading (GDPval), and image data and evaluation (ChatGPT Images). The vendors behind those campaigns are OpenAI's standard set — reported to include Mercor, Surge and Handshake — but her specific vendor relationships are inference.

Career

  1. presentProgram Manager, Human Data, OpenAI — Headline 'OpenAI | Human Data'; posts the team's own PM job openings

On the record

Human Data team · 2025-11

Recruits publicly for OpenAI's Human Data team, calling it one of the best teams in the org to join and the work genuinely fun. source

Hiring · 2025-11

Circulates the team's Program Manager, Human Data opening and asks people to pass it on. source

Recent posts

Alexandra Barr
OpenAI | Human Data
Writing is really hard. So I'm trying to do it more I think good things come if you believe it's okay to embarrass yourself in the pursuit of human connection Here's my attempt to imperfectly express myself and practice the act of craft/creation. Join me, if you like :) poastoffice.com
10010 comments
Alexandra Barr
OpenAI | Human Data
OpenAI Human Data crew just landed at ICLR! Catch us at the GDPval poster session (booth 1819) at 3pm today!
2286 comments
Alexandra Barr
OpenAI | Human Data
I've been having so much fun trying out the presets on ChatGPT Images! Almost every single one of my teammates has turned into a holiday-themed character on Slack ✨❄️ Congratulations to the whole team — this was a huge team effort! Grateful to have contributed to the Data & Evaluation work on this release and beyond excited to see the incredible images everyone’s about to dream up :) Check out the full blog post here: lnkd.in/e-d4mgAq
14315 comments • 1 reposts
Alexandra Barr
OpenAI | Human Data
In celebration of 10 years, OpenAI has 10 easter eggs in our merch drop at supply.openai.com My personal favorite is the Thinking Deeply hat :) check it out, if you like!
28

Interviews & talks

Why it matters here

High. She is inside the image data and evaluation group, she works on expert-graded evals, and she is publicly reachable — a rare combination. For a creative/taste-data vendor she is the person most likely to actually read a cold approach at OpenAI's human-data function and route it to whoever owns budget. She is not the economic buyer; treat her as the way in.

How to reach

X is the live channel — she posts there and engages. Open with something specific to image evaluation rather than a capability list: e.g. that rubric-based grading works for correctness but degrades on aesthetic and compositional judgement, and that Julian has a qualified panel of designers and art directors who can produce graded preference data on exactly that. Offer a free small calibration set rather than a meeting. Speculation: given she recruits publicly, an approach framed as 'here is a supply pool you cannot hire in-house' will read better than a vendor pitch.

What we could not establish
  • Exact title and level unconfirmed — she describes herself only as 'OpenAI | Human Data'; 'Program Manager' is inferred from the role she posts and from OpenAI's team structure.
  • No education or pre-OpenAI career found.
  • OpenReview profile exists but was behind a browser check, so its contents are unverified.
  • Whether she still works on image data in 2026 (latest hard credit is Dec 2025).

Why they are on the map

Listed under 'Data & Evaluation' in ChatGPT Images (gpt-image-1.5) credits (Dec 2025); GDPval co-author (Oct 2025); LinkedIn headline 'OpenAI | Human Data' source — evidence dated 2025-12

Runs vendor projects; decides if a pilot converts.

How OpenAI buys data

Has a dedicated Human Data team that 'turns human feedback into reliable signals for training and evaluation' via bespoke campaigns with 'external vendors and AI trainers' and has open PM/TPM/EM/TLM roles on it (2025-26 job posts). Ended its Scale AI relationship after Meta's stake (Jun 2025); reported to be a major revenue source for Mercor and to use Surge and Handshake, while building in-house data capacity. GDPval (Oct 2025) and HealthBench (May 2025, 262 physicians) show it buys large expert-graded eval sets. Image models ship with a named 'Data & Evaluation' team (Dec 2025 credits).

A central Human Data team (PMs, TPMs, engineers) works with all research teams and vendors. Glaese, a former Human Data lead, is now VP Research & Safety (Jul 2026). Image generation sits under Multimodal (Dhariwal), with its own data leads and an 18-person Data & Evaluation group, plus Model Behavior. Sora was shut down in Mar 2026 and Peebles left in Apr 2026. Image work was folded into GPT Image 2 / 2.5.

Known vendors: mercor, surge-ai, handshake-ai, scale-ai (ended Jun 2025)

Sources

  1. New ChatGPT Images credits (Data & Evaluation)
  2. Introducing deep research credits (Foundational Contributors)
  3. GDPval paper author list
  4. Her X post recruiting for the OpenAI human data team
  5. LinkedIn profile 'OpenAI | Human Data'