Miju Labs

← All buyers · Black Forest Labs

P

Patrick Esser

Co-founder · Black Forest Labs

medium confidenceTier A · Talk to firstImage, video & audio modelsimage · video

Esser is the third founding researcher, co-author of VQGAN ('Taming Transformers') and of the Stable Diffusion and SD3 papers. Of the three founders he is the one who has spoken most directly about taste: he has publicly framed curation of samples as something the model cannot do for you. That is a rhetorical opening a creative-data vendor can build on.

Background

Esser worked with Björn Ommer at Heidelberg and LMU Munich, where 'Taming Transformers for High-Resolution Image Synthesis' (VQGAN) came out of the group and became a foundation of the text-to-image era. He joined Stability AI, was a lead author on Stable Diffusion 3's rectified-flow scaling paper, and co-founded Black Forest Labs in 2024.

He is the least publicly visible of the three founders. His most quotable public appearance remains the August 2024 a16z podcast the founders did together around the FLUX.1 launch.

What they run now

Core model research across the FLUX line. His public framing of the product problem is about controllability and prompting: he has said they trained the model to respond well to varied prompting techniques, and that the human remains the one who decides which of a hundred generations is right for a project. Inside BFL that maps to prompt-following rubrics, editing fidelity and the sample-selection loop — all of which are rubric-and-rater problems.

What they buy

Inference. Prompt-following and editing-fidelity rubrics, expert pairwise preference data for image editing (the Kontext line), and human-judged sample selection for SFT sets. No named vendor. FLUX.1 Kontext shipped with a crowd-sourced KontextBench, which shows willingness to use external raters at least once.

Career

  1. 2024 – presentCo-founder, Black Forest Labs
  2. 2021 – 2024Researcher (Stable Diffusion, SD3), Stability AI
  3. 2017 – 2021PhD researcher, Heidelberg University / LMU Munich (Ommer group) — VQGAN / Taming Transformers
  4. c. 2017–2021Heidelberg University, PhD, computer vision (Ommer group)

On the record

taste is human · 2024-08

Says taste and curation of the sample thread stay with the person: you can make a hundred samples fast, but which one serves your project is your call. source

prompt robustness · 2024-08

Notes they deliberately trained the model to react well to a range of prompting techniques. source

Interviews & talks

Why it matters here

Moderate on his own, high as a co-signer. His public line that taste stays with the human is the exact premise of Julian's business, and quoting it back to BFL is an honest, non-flattering way to open a conversation with any of the three founders.

How to reach

Speculation: use his taste quote as the hook in outreach to BFL generally rather than targeting him alone — he is the least reachable founder and has no found public handle. If a route opens, the substantive offer is an editing-fidelity eval on Kontext-style tasks (instruction adherence, identity preservation, unwanted-change detection) scored by retouchers and photo editors.

What we could not establish
  • No confirmed LinkedIn or X handle found.
  • Which BFL workstream he leads day to day.
  • Whether KontextBench's crowd raters came through a vendor.

Why they are on the map

BFL co-founder (latent diffusion/rectified-flow author); FLUX.1 Kontext author. Exact current title not verified. source — evidence dated 2025-06

Needs the data for a model; writes the spec.

How Black Forest Labs buys data

FLUX 3 launch (2026-07-23) reported head-to-head human preference results vs Luma Ray 3.2 (93%), Runway Gen-4.5 (77%) and Grok Imagine Video; raters and methodology not disclosed. Open role 'Data Procurement & Operations Lead' (Freiburg/SF, posted around Jan 2026) will own data vendors, specialist contractors, stock media and licensors, and build 'collection and evaluation systems through internal tools, vendor partnerships'. 'MTS - Post Training' role (2026) covers RLHF/DPO/reward modeling and 'human prefs'. FLUX.1 Kontext (2025-06) introduced crowd-sourced KontextBench.

Flat 'Member of Technical Staff' research org led by founders. Teams visible in hiring: Pretraining, Post Training, VLM, Image/Video Generation, Research Infra. Data procurement/ops is being set up as its own lead role, and no named holder was found.

Sources

  1. Data Procurement & Operations Lead — Black Forest Labs (Greenhouse)
  2. The AI startup that has quietly become one of Europe's most valuable AI companies (Fortune, 2026-02-17)
  3. FLUX 3 — Real World Models (BFL blog, 2026-07-23)
  4. Black Forest Labs on FLUX — a16z AI Podcast transcript (2024-08-18)
  5. Flux (text-to-image model) — Wikipedia
  6. Black Forest Labs: Europe's most-hyped — and elusive — startup? (Sifted, 2025-04-24)