Black Forest Labs
0 open Forward Deployed Engineer roles at Black Forest Labs, plus 2 closed. Curated with comp bands.
On the board since Jun 2026 · latest role Jul 2026 · no roles open right now
Our take
Why Black Forest Labs
The FLUX lab's forward-deployed seat has shifted from image inference to robotics — a $3.25B frontier media lab now embedding engineers on customers' actual robots.
- FDE type
- True forward-deployed
- Travel
- Medium
- Comp
- Competitive
- Stage
- Frontier lab
A good fit if you're…
- ML engineers who know diffusion and flow-matching internals, not just API wiring
- Engineers excited by physical AI and robotics on top of generative-media models
- People who want a frontier-lab seat outside the LLM crowd, near-zero candidate competition
- Candidates open to on-site work in Freiburg, Germany
A frontier media lab now worth $3.25B
Black Forest Labs builds FLUX, among the most widely used open image-generation model families — integrated by Adobe, Canva, Meta, Microsoft, Vercel and ElevenLabs. In December 2025 it raised a $300M Series B at a $3.25B valuation, backed by Salesforce Ventures, a16z and NVIDIA. It is the clearest example of the frontier-lab playbook applied outside language models.
The forward-deployed seat moved to robotics
The Forward Deployed Machine Learning Engineer role that defined this entry has closed. The live forward-deployed seat is now a Forward Deployed Robotics Engineer, on-site in Freiburg, Germany: you embed with BFL's robotics and physical-AI customers to design, prototype and ship integrations and finetunes of BFL models on real robots, then feed learnings back to the research team.
Why FLUX 3 makes this a physical-AI job
On July 26, 2026 BFL released FLUX 3, a single flow-matching backbone spanning image, video, audio and robot action prediction — including FLUX-mimic, a robot policy that runs in under 80ms on one RTX 5090. The forward-deployed robotics seat sits directly downstream of that research bet: the lab now treats physical dynamics as just another modality to deploy.
Comp: a real German band, equity on top
The live robotics seat lists EUR 120K-180K, on-site in Freiburg — a concrete band rather than the blank it used to be. As a frontier lab, expect meaningful equity layered on cash. The specialized diffusion and flow-matching requirement is unusual leverage: a qualified candidate is negotiating from a very small pool.
The skills that actually gate this
This is ML-heavy, not generic integration. BFL wants diffusion-model and flow-matching knowledge plus transformer inference-optimization experience, and the ability to tune latency and output quality across on-prem, hosted, real-time and edge deployments. That requirement shrinks the candidate pool to near zero, which is exactly why it is worth knowing about.
Who thrives, and who bounces
Thrives: ML-deep engineers who enjoy customer-facing deployment and physical systems, and who can relocate to or work on-site in Freiburg. Bounces: generalist integrators without model internals, anyone who needs US-remote work, or candidates expecting an LLM-centric frontier lab — this is generative media and robotics, not language models.
Questions candidates ask
Is this an ML role or an integration role?
ML-heavy. It needs diffusion and flow-matching internals plus inference-optimization experience, not generic API integration work.
Where is the live seat based?
Freiburg, Germany, on-site. The earlier Forward Deployed Machine Learning Engineer seat has closed; the live forward-deployed role is a Robotics Engineer.
What does it pay?
The robotics seat lists EUR 120K-180K, with equity on top as a frontier lab. The specialized requirement gives strong candidates real leverage.
Why robotics at an image lab?
FLUX 3 added robot action prediction to the same backbone, so BFL now treats physical dynamics as another modality — and is embedding engineers on customers' robots to deploy it.
Analysis updated 2026-10-01
Open roles at Black Forest Labs
Compare Black Forest Labs
Head-to-head
Others in the same tier
Free forever. Unsubscribe anytime.