Fireworks Developer Advocate at Fireworks builds developer-focused tools and content to promote Fireworks' AI platform. The role requires shipping code and public technical communication to engage the developer community.
Responsibilities
Developers decide what infrastructure wins. They try three platforms in an afternoon, keep the one that got them to a working inference call fastest, and tell everyone else about it. As a Developer Advocate at Fireworks, you own a piece of that decision: you make Fireworks the platform developers reach for first when they want to build, fine-tune, and serve frontier open models in production. You do that by building things worth showing, not by talking about things other people built.
This role has a dual bar, and we mean both halves of it. You are a working engineer: you ship code, read diffs, deploy models, and can hold your own in a conversation about batching, quantization, or why a fine-tune regressed. You are also a public technical communicator: you write clearly enough that a senior engineer forwards it, you're good on a stage, and you have or want a real presence in the places developers actually argue about this stuff. Most people are strong at one and passable at the other. We're looking for the rarer profile who is genuinely credible at both, because in AI infrastructure the audience can tell within one paragraph whether you've actually run the thing.
The field moves weekly. A major open-weights model drops and the window to be the definitive resource on it is measured in days. Your job is to be one step ahead of that cadence, consistently, and to make it look repeatable rather than heroic.
This is one of the first advocate seats on the team, which means you'll help set the standards: for content, for launch response, and for how we show up. We're hiring at either the Senior or Staff level and will determine level through the interview process based on your experience.
Location and Work Style
This role is based in San Mateo, CA or London. You'll travel regularly to the conferences, hackathons, meetups, and partner events where the developer community shows up. Expect roughly 20% to 25% travel in a typical quarter, concentrated around major events and launches.
If you're based in London, you'll anchor our presence in the European developer community, work across time zones with a primarily US-based team, and travel periodically to our San Mateo HQ for onsites and planning.
Build things developers can run. Ship cookbooks, quickstarts, reference applications, and end-to-end tutorials that take a developer from zero to a working inference call and then to something production-shaped. Everything you publish is code someone can fork and run, not a screenshot of code that worked once on your machine. The bar is content a senior engineer respects and a new developer can follow, ideally both in the same artifact.
Own the launch response. When a significant model or technique lands, you're among the first credible voices with a working demo, an honest benchmark, and a "how to actually use this today" guide. You'll help build the internal loop (model access, eval harnesses, publishing paths) that makes day-one content a process rather than an all-nighter.
Show up where developers are, owned and unowned. On our properties, that's docs, blog, examples, cookbooks, Discord, and the developer newsletter. On properties we don't own, that's GitHub, X, Reddit, YouTube, Hacker News, and the open-source projects our developers depend on, including contributing code and reviewing PRs where it matters. You answer real questions in public and turn the recurring ones into permanent content.
Speak, teach, and run the room. Conference talks, hands-on workshops, hackathon mentoring, and developer dinners. You'll help decide where Fireworks shows up and what we say when we get there, and you'll build the material so the next person can deliver it without starting over.
Be the developer's voice inside Fireworks. Bring friction, feature gaps, and unmet needs back to product and engineering with evidence: a failing repro, a thread of five people hitting the same wall, a competitor's ergonomics that are simply better. Then close the loop publicly when we ship the fix.
Qualification
You can code and read diffsHands-on with the modern AI stackA public track recordDemonstrated speaking abilityLearning velocityOwnership instinctDepth in inference optimizationComfort making videoSolve Hard Problems
Required
You've spoken at multiple significant events and left behind a workshop or talk that others on the team can now deliver.
Product and engineering can point to specific roadmap decisions that changed because of feedback you surfaced.
The content and launch standards you set are what new advocates onboard into, and they hold up without you in the room.
5+ years of combined engineering and developer-facing experience, with real time in both. For example, several years shipping production software plus meaningful time in DevRel, developer education, technical writing, or solutions engineering. We're flexible on the ratio and inflexible on the requirement that both are genuinely there.
You can code and read diffs. Strong Python; comfortable in TypeScript/JavaScript. You've built and deployed applications yourself, and you can review someone else's sample and tell whether it's good.
Hands-on with the modern AI stack. You've worked with LLM APIs in production, built with at least one agent framework, and have personally fine-tuned, evaluated, and deployed a model, not just read about it. You understand what inference latency, throughput, and cost actually depend on.
A public track record. Technical writing, talks, videos, open-source work, or a following you earned. Something we can go read or watch that shows you can explain hard things well to engineers.
Demonstrated speaking ability. You've presented at conferences, meetups, or workshops, and you're comfortable doing it on short notice about something that shipped last week.
Learning velocity. You can go from "this paper came out Tuesday" to "here's a working implementation and an honest assessment" quickly, and you have a system for staying current rather than relying on luck.
Ownership instinct. You pick your own topics, scope your own projects end to end, and don't need a content calendar handed to you. This role is early enough that you'll be defining the work as often as executing it.
Preferred
Open-source contributions or maintainership in the AI/ML ecosystem, including inference engines, model libraries, agent frameworks, and evaluation tooling.
Depth in inference optimization, serving architecture, post-training, or evaluation methodology.
You've built something that outlived your direct involvement: a content operation, a launch process, an events playbook, or a community program.
Comfort making video.
Existing relationships with AI research or open-model communities.
You use frontier coding agents in your daily work and have opinions about where they help and where they don't.
Experience being early on a DevRel team and helping establish how it operates.
Solve Hard Problems: Tackle challenges at the forefront of AI infrastructure, from low-latency inference to scalable model serving.
