What if the hardest part of generative AI isn't training the model, but making it work in production under constraints no one anticipated?
Our founding team pioneered Latent Diffusion and Stable Diffusion - breakthroughs that made generative AI accessible to millions. Today, our FLUX models power creative tools, design workflows, and products across industries worldwide.
Our FLUX models are best-in-class not only for their capability, but for ease of use in developing production applications. We top public benchmarks and compete at the frontier - and in most instances we're winning.
If you're relentlessly curious and driven by high agency, we want to talk.
With a team of ~50, we move fast and punch above our weight. From our labs in Freiburg - a university town in the Black Forest - and San Francisco, we're building what comes next.
What You'll Pioneer
You'll live at the intersection of cutting-edge research and brutal production reality. Your customers won't just want FLUX to work—they'll need it optimized for their specific hardware, fine-tuned for their unique use cases, and integrated into systems that weren't designed for diffusion models in the first place.
You'll be the person who:
- Ensures FLUX models perform optimally in customer environments—whether that's on-premise GPU clusters or BFL-hosted infrastructure—balancing the eternal tension between latency and output quality
- Architects deep product integrations that go far beyond "here's an API endpoint"—helping customers with everything from model hosting and deployment to inference optimization techniques that haven't made it into textbooks yet
- Customizes our foundation models for visual media to solve problems customers couldn't articulate until you helped them understand what's possible
- Sits in technical deep-dives with customers to diagnose performance bottlenecks, then translates those findings into solutions (and sometimes into research questions for our core team)
- Discovers where generative visual AI should go next by understanding what industries are struggling with problems we could solve
Questions We're Wrestling With
- What does "optimal performance" actually mean when one customer needs 100ms latency and another needs photorealistic quality at any cost?
- How do you fine-tune a foundation model for a customer's specific use case without losing what made it powerful in the first place?
- When should a customer run FLUX on their own infrastructure versus use our hosted solution—and how do we help them make that decision honestly?
- What inference optimizations work in theory but break in production, and vice versa?
- Which industries don't yet realize they have a generative visual AI problem we could solve?
We're figuring these out together, at the edge of what's technically possible.
Who Thrives Here
You understand diffusion models not just conceptually, but viscerally—you've debugged them, optimized them, served them at scale. You've been in the room when a customer's integration goes wrong and you need to diagnose whether it's a model issue, an infrastructure issue, or a fundamental misunderstanding of what the model can do.
You likely have:
- Direct experience working with customers on generative AI deployment—the kind where you're iterating on solutions in real-time, not just following a playbook
- Hands-on expertise with generative modeling approaches, particularly finetuning, optimizing, and serving deep learning models in production environments
- A proven track record as an ML engineer who's shipped models that real systems depend on
- Strong Python skills and intuitive understanding of API design (because demos and prototypes are how you communicate what's possible)
- The ability to explain why a diffusion model is slow to an executive and how to fix it to an engineer—in the same meeting
We'd be especially excited if you:
- Have deep knowledge of diffusion models and/or flow matching, including finetuning and distillation techniques that go beyond standard tutorials
- Know the FLUX ecosystem intimately—ComfyUI, common training frameworks, the tools practitioners actually use
- Have battle-tested experience optimizing inference for transformer-based models (and the scars to prove it)
- Can architect solutions in complex enterprise environments where "just add more GPUs" isn't an option
- Contribute to open-source projects in the diffusion model space and understand the community
- Have deployed models on cloud platforms using state-of-the-art serving infrastructure
What We're Building Toward
We're not just supporting customers—we're learning what it actually takes to bring frontier generative AI into production at scale. Every customer integration teaches us something we didn't know. Every optimization challenge reveals gaps in our understanding. If that sounds more compelling than having all the answers documented, we should talk.
How We Work Together
We’re a distributed team with real offices that people actually use. Depending on your role, you’ll either join us in Freiburg or SF at least 2 days a week (or one full week every other week), or work remotely with a monthly in-person week to stay connected. We’ll cover reasonable travel costs to make this possible. We think in-person time matters, and we’ve structured things to make it accessible to all. We’ll discuss what this will look like for the role during our interview process.
Everything we do is grounded in four values:
- Obsessed. We are a frontier research lab. The science has to be right, the understanding deep, the product beautiful.
- Low Ego. The work speaks. The best idea wins, no matter who said it. Credit is shared. Nobody is above any task.
- Bold. We take the ambitious bet. We ship, we do not wait for conditions to be perfect.
- Kind. People over politics. We treat each other with genuine warmth. Agency without empathy creates chaos.
Base Annual Salary for SF based role: $180,000–$300,000 USD (depending on experience)
We're based in Europe and value depth over noise, collaboration over hero culture, and honest technical conversations over hype. Our models have been downloaded hundreds of millions of times, but we're still a ~50-person team learning what's possible at the edge of generative AI.
