Senior Developer Advocate

Based in the San Francisco Bay Area, or willing to relocate. 

The role

We're looking for a scrappy, resourceful Developer Advocate who understands inference to help developers run open models in production on Token Factory, Nebius' high-performance inference platform. 

Every agent, copilot and AI product runs on an inference layer, and the developers choosing that layer have hard questions: Which open model should I run for this step? What does this cost per million tokens at my traffic? Why is my p99 latency spiking? Why does my multi-turn agent get slower and more expensive as the context grows? When do I move from a serverless endpoint to a dedicated one, and when does fine-tuning beat a longer prompt? Your job is to have credible, hands-on answers to those questions, and to show the work in public. 

You'll be embedded with the Token Factory product and engineering team in San Francisco. You'll be in their standups and bug bashes, you'll carry customer stories back to them, and you'll fix the docs gap yourself when you find one. You'll be equally at home in the Bay Area's AI-native community: the meetups our customers and partners host, and the open-source inference projects developers actually use. 

This is a hands-on, builder-first role. The content that matters here is less "how to call the API" and more "here's how a production inference stack is put together, and here are the tradeoffs." If you have strong opinions about serving open models and want a platform to prove them on, this role is for you. 

This role is based in the San Francisco Bay Area. The Token Factory team's center of gravity is in San Francisco and we expect this person to be part of the local community in person. 

Your responsibilities will include:

• Help developers and teams discover, evaluate and adopt Token Factory for production inference workloads, from a first API call on a serverless endpoint to dedicated endpoints and fine-tuned models. 

• Build demos, benchmarks and reference architectures that make real tradeoffs concrete: which open model to use for which step, latency against cost per token, how context growth affects multi-turn agents, reliability of tool calling and structured outputs, and when fine-tuning or a custom speculator pays off versus a bigger model or a longer prompt. 

• Show how Token Factory fits into the stack developers already use: OpenAI-compatible SDKs, agent frameworks, MCP and tool use, evals, embeddings and retrieval, and post-training. 

• Be embedded with Token Factory product and engineering: join standups and bug bashes, test new models and features before launch, and make sure the right samples and docs exist on day zero of every release. 

• Own the developer feedback loop. Gather what's breaking for builders, bring it back as concrete product input, and follow through until it ships or is explicitly declined. 

• Represent Nebius in the Bay Area inference and open-model community: meetups, partner and customer events, open-source communities such as vLLM, SGLang and Ray, and developer conferences. 

• Publish technical content and give talks that take developers from first hearing about Token Factory to running something on it, and partner with marketing to turn real builder stories into case studies. 

• Work closely with the DevRel team and the Token Factory product marketing team on launches, community programs and content strategy. 

We expect you to have: 

• Hands-on experience running open models in production or at scale through an inference provider or a serving stack, with a clear point of view on how you chose it and what broke. If you haven't tried Token Factory yet, we'll expect you to have done so before we talk. 

• A working understanding of what happens under the hood of a serving engine like vLLM or SGLang: continuous batching, KV and prefix caching, quantization, speculative decoding, and how each shows up in latency and cost. You won't be tuning these on Token Factory, but you need to hold your own with the engineers who do. 

• Current, first-hand knowledge of the open model ecosystem, including which models are worth recommending for which workloads and why. 

• Comfort writing and reviewing code, reproducing issues, and shipping fixes to docs and samples yourself. 

• A track record of building and explaining in public: talks, open-source contributions, writing, or an active technical presence on GitHub, X or LinkedIn. A DevRel title is not required. 

• The ability to explain complex systems clearly to technical and non-technical audiences, and to be a credible technical partner to engineers, product managers and customers. 

• Based in the San Francisco Bay Area, or willing to relocate. 

Candidates who come from forward-deployed engineering, solutions architecture, ML engineering or technical product roles are strongly encouraged to apply. 

It's a plus if you have: 

• Hands-on experience with supervised fine-tuning, distillation or RL post-training of open models, and a view on when it beats a longer prompt or a bigger model. 

• Contributions to open-source inference or distributed-compute projects such as vLLM, SGLang or Ray. 

• Experience at an inference or GPU cloud provider, or as a heavy user of several. 

• An existing audience that trusts your takes on models and infrastructure. #LI-LT1

Key employee benefits in the US:

• Health insurance: 100% company-paid medical, dental, and vision coverage for employees and families.
• 401(k) plan: Up to 4% company match with immediate vesting.
• Parental leave: 20 weeks paid for primary caregivers, 12 weeks for secondary caregivers.
• Remote work reimbursement: Up to $85/month for mobile and internet.
• Disability & life insurance: Company-paid short-term, long-term and life insurance coverage.