Miju Labs

← All buyers · OpenAI

Phoebe Thacker

Global Head of Data Research Programmes & London Site Lead · OpenAI

medium confidenceTier A · Talk to firstFrontier labstext

Thacker runs OpenAI's human-data research programmes globally and is OpenAI's London site lead. She is the most senior operator publicly identifiable on the human-data side at OpenAI: the person whose organisation designs the campaigns, recruits and manages the expert annotator pools, and contracts the vendors that produce OpenAI's training and evaluation data. She was a co-author on GDPval, OpenAI's expert-graded real-work benchmark.

Background

Public material on her is thin before OpenAI. She surfaces first as a data-operations leader inside OpenAI's research organisation, based in London, and by 2026 is described in press coverage of OpenAI's UK expansion as 'Global Head of Data Research Programmes and London site lead'. Her LinkedIn headline in 2026 reads 'Head of Human Data Operations' / 'Research Data Ops @ OpenAI', which is the same job described from the operations rather than the programme side.

Her dual mandate matters: she is simultaneously the data-programmes owner and the site lead for OpenAI's largest research hub outside the US (~200 staff in London, roughly 30 of them researchers, moving into an 88,500 sq ft Regent Quarter office in King's Cross in 2027). That puts European expert-labour supply inside her natural remit.

What they run now

She owns the programme layer of OpenAI's human data: how research teams' data needs are turned into running campaigns with external vendors and AI trainers, how expert pools are recruited and qualified, and how quality is measured. GDPval (Oct 2025) is the clearest public artefact of that work — 1,320 real occupational deliverables graded by experienced professionals across 44 occupations — and it is exactly the kind of programme her function stands up. As London site lead she also controls headcount and hiring for the UK hub, which is where OpenAI's European expert-recruitment story would sit.

What they buy

Expert human data programmes at scale: SFT and preference data, expert-graded evaluation sets, and the annotator/expert recruitment behind them, across every modality OpenAI trains on. GDPval shows the shape of what her org commissions — deliverable-level work products from experienced practitioners, plus expert grading of model output against them. OpenAI's public description of the Human Data team says it runs bespoke campaigns with 'external vendors and AI trainers', and OpenAI is reported to use Mercor, Surge and Handshake after ending its Scale AI relationship in mid-2025. Which vendors she personally signs is inference, not established fact.

Career

  1. 2026 – presentGlobal Head of Data Research Programmes; London Site Lead, OpenAI — Title as given in April 2026 UK-expansion press coverage
  2. presentHead of Human Data Operations / Research Data Ops, OpenAI — LinkedIn headline; same function, earlier framing

On the record

UK talent · 2026-04

Says the UK has 'incredible depth of talent and a strong track record in AI', and that London is already a key hub for OpenAI research. source

Interviews & talks

Why it matters here

She is the single highest-leverage human-data contact at OpenAI that Julian can name from public sources, and she is in Europe. A creative/taste-data specialist selling image, video and design judgement is selling exactly the category her programmes commission: expert-graded deliverables and preference data. Her London base also means a Lübeck-based vendor is in the same timezone and the same labour market she is trying to recruit from, which is a real, non-generic reason for her to take a call.

How to reach

Open with GDPval, not with a pitch: the honest observation that GDPval covers occupational deliverables but creative/design work is the hardest category to grade with rubrics, and that Julian has built a graded pool for exactly that. Concretely propose a small paid pilot: a European panel of practising designers/art directors producing deliverables plus rubric-based grading of model output on visual and layout tasks. LinkedIn is the right channel (public profile); X is secondary. Speculation: given her site-lead role, an angle on European creative-professional supply (Germany, UK, Nordics) is likely to land better than a generic capacity pitch.

What we could not establish
  • No verified education or pre-OpenAI career history found.
  • Two LinkedIn vanity URLs appear in search results (/in/phoebe-thacker/ and /in/phoebe-thacker-0a869927/); could not confirm which is canonical.
  • The X handle @phoebethacker matches her name and reposts OpenAI launch announcements, but X is robots-blocked so I could not read a bio naming OpenAI — treat the handle as probable, not confirmed.
  • Whether 'Global Head of Data Research Programmes' and 'Head of Human Data Operations' are the same role renamed, or two roles, is unclear.
  • No evidence on which vendors she contracts with personally.

Why they are on the map

GDPval co-author (Oct 2025); role unverified source — evidence dated 2025-10

Owns the human-data budget and vendor list.

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. OpenAI to move to first permanent London office (Thacker quoted with title)
  2. Why OpenAI is embarking on a major UK expansion
  3. GDPval paper (author list includes Phoebe Thacker)
  4. StartupHub profile listing her title