Software Engineer, Fireworks AI

• 1 - 7 years of software engineering experience (We are hiring at multiple levels for this role)
• Passion for Developer Experience: You care deeply about API design, documentation, and the ergonomics of CLI/SDK tools
• Understanding of the GenAI Lifecycle: You understand the end-to-end workflow—from prompting a base model to curating a dataset, fine-tuning, and productionizing agents—and how these steps interconnect
• User-Centric Mindset: Willing to talk to users, triage GitHub issues for open-source projects, and build products from scratch to serve emerging needs
• Strong communication skills: Ability to explain why a specific evaluation metric matters to a customer and how it impacts their business logic

Desirable

• We are seeking a Software Engineer, DevEx & Evals to play a highly impactful role in shaping the Fireworks platform
• You will be responsible for defining and building a cohesive developer journey that bridges the gap between experimentation and production
• In the AI application lifecycle, Evals and Training is a continuous loop
• You will build the experiences that allow developers to seamlessly navigate this cycle: starting with serverless experimentation, moving to fine-tuning custom models, evaluating them rigorously, and finally deploying to on-demand GPUs
• You will also lead the engineering efforts on our open-source initiative and productize these capabilities to help users author evals and train custom models
• As the public platform serves as the product-led growth (PLG) engine for Fireworks, your work will directly drive business impact by removing friction at every stage of the developer onboarding and adoption process
• Unify the AI Lifecycle: Build a streamlined experience that connects our serverless inference, fine-tuning, and on-demand deployment products into a single, intuitive workflow
• Streamline Developer Onboarding: Obsess over the "Time to First Token" and "Time to First Fine-tune." Identify and eliminate friction points for new developers entering the platform
• Architect Scalable Eval Tooling: Design full-stack features that support the continuous cycle of training and evaluation, providing deep insights into model quality that directly inform the next round of fine-tuning
• Productize Inference Optimization: Build experiences that help developers optimize their GPU deployments for their specific workloads, guiding them on the right balance of throughput, latency, and model quality
• Data-Driven Prototyping: Rapidly prototype new ways to visualize evaluation data and training metrics, helping users make sense of complex benchmarks