Miju Labs

← All buyers · Black Forest Labs

Robin Rombach

Co-founder & CEO · Black Forest Labs

high confidenceTier A · Talk to firstImage, video & audio modelsimage · video · multimodal

Rombach is the first author of the Latent Diffusion / Stable Diffusion line and now runs Black Forest Labs from Freiburg im Breisgau, roughly a two-hour drive from Julian in Lübeck's sense of 'domestic'. He is the final signature on what BFL spends on data and evaluation, and he personally frames data pre-processing and filtering — not architecture — as the place a visual lab distinguishes itself. FLUX 3 shipped in July 2026 with headline human-preference win rates and no disclosed rater methodology, which is precisely the gap a credible creative-eval vendor fills.

Background

Rombach did his PhD at LMU Munich in Björn Ommer's computer-vision group, where he was lead author on 'High-Resolution Image Synthesis with Latent Diffusion Models' — the paper that became Stable Diffusion. He went to Stability AI with Andreas Blattmann and Patrick Esser, led the Stable Diffusion XL and SD3 work, and left in 2024 with them to found Black Forest Labs.

BFL raised a ~$35M seed/Series A in 2024 and then $300M at a $3.25B valuation in late 2025, co-led by AMP and Salesforce Ventures with Canva, Figma, a16z and Nvidia participating. Rombach has kept the company deliberately quiet and deliberately German-based, and told Fortune he expects FLUX to have changed the trajectory of visual AI by 2028. Martin Scorsese was announced as an advisor in June 2026.

What they run now

Rombach owns BFL's model roadmap and its commercial shape. Since July 2026 that means FLUX 3, a joint image/video/audio/action flow model, plus FLUX.2 (Nov 2025), the Kontext editing line, and vertical products (FLUX Erase, FLUX VTO, a 2K/4K video upscaler) sold into Meta, Canva, Figma, Envato and Adobe-adjacent workflows. Revenue splits roughly 50/50 between usage-driven API products and enterprise licensing. The company has never disclosed its training-data sources and has denied using LAION, which makes licensed and commissioned data a standing commercial need rather than a research nicety.

What they buy

Inference, but well-grounded. BFL is standing up a formal data-procurement function (see the open Data Procurement & Operations Lead role, Freiburg/SF) whose remit explicitly covers data vendors, stock media providers, specialised contractors and direct content licensors, plus building 'quality evaluation frameworks'. The FLUX 3 launch reports pairwise human-preference win rates against Runway Gen-4.5 (77%), Luma Ray 3.2 (93%), Grok Imagine, Kling v3 Pro and Gemini Omni Flash, with raters, counts and rubric undisclosed and the results self-described as preliminary. So: licensed image and video corpora; expert pairwise preference data for image, video editing and now audio/action; and an evaluation panel that can be cited as independent. No vendor has been named publicly.

Career

  1. 2024 – presentCo-founder & CEO, Black Forest Labs — Freiburg im Breisgau, plus San Francisco
  2. 2022 – 2024Research lead (Stable Diffusion XL, SD3), Stability AI
  3. 2019 – 2022PhD researcher, LMU Munich (Ommer group) — Lead author, Latent Diffusion Models
  4. 2019–2022Ludwig Maximilian University of Munich, PhD, computer vision / machine learning

On the record

where labs differentiate · 2024-08

Says the differentiating work is scaling training infrastructure and doing data pre-processing and filtering well, rather than novel architecture. source

human judgement in training · 2024-08

Argues the ability to judge where a training run is heading from early samples is extremely important — an explicitly human, taste-based skill. source

editing as the unlock · 2026-02

Describes FLUX.1 Kontext as the first model able to edit images while holding character consistency. source

Interviews & talks

Why it matters here

He is the single highest-value name in this batch. Geography is real leverage — Freiburg is a same-day train from anywhere in Germany, and a German vendor who can run rater panels under EU contracts and GDPR is a structurally easier procurement than a US crowd platform. BFL publishes preference numbers it cannot fully defend; an independent, documented expert-rater evaluation of FLUX 3 against Ray 3.2 and Gen-4.5 is a product Julian could scope in a week and that Rombach's own launch post creates demand for.

How to reach

Do not pitch labelling. Pitch evaluation credibility. Speculation: the strongest cold open is a short, unpaid, fully methodologised mini-eval — 200 prompts, a published rubric, named rater qualifications — comparing FLUX 3 to one named competitor on the dimensions BFL already claims, sent with the note that the launch post's win rates have no disclosed methodology. Route in via the Data Procurement & Operations Lead role if it is filled by then; otherwise via Rombach or Axel Sauer directly on X, in German or English. A Freiburg visit offer is cheap and differentiating.

What we could not establish
  • Whether the Data Procurement & Operations Lead role has been filled — as of 2026-09-22 the Greenhouse posting is still live and no holder is publicly named.
  • Who actually ran the FLUX 3 preference evaluations (internal staff, contractors, or a vendor).
  • Whether BFL licenses video corpora through named partners, and who negotiates those.

Why they are on the map

CEO quoted at the FLUX 3 launch, which was announced with human-preference win rates; final say on data/eval spend. source — evidence dated 2026-07

Decides whether the problem is funded.

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)