About Us:
Fireworks is the platform for specialized intelligence, enabling companies to build, train, and serve AI models tailored to their own data, workflows, and products. Founded by the team behind PyTorch and backed by AMD, Atreides, Benchmark Capital, Index Ventures, Lightspeed, NVIDIA, Sequoia Capital, and TCV, Fireworks powers production AI with hundreds of state-of-the-art open models across text, image, embedding, audio, and multimodal workloads. Today, Fireworks is a Series D company valued at $17.5 billion, bringing together an ambitious, collaborative team that's building the future of enterprise AI.
The Role:
GPUs are the most expensive and supply-constrained input in AI. Our ability to effectively deploy them towards our customer demand is one of the critical inputs to our growth. We're hiring our first Data Scientist to build and own the quantitative models that drive pricing, packaging, and capacity allocation in order to optimize capacity and balance supply with demand.
The ideal candidate brings expertise at the intersection of operations research and economics with the ability to both model our market, build dynamic pricing models, and design experiments and causal studies. This is a hands-on role: you'll write production-quality code, ship models that drive real decisions, and bring an true ownership mentality to help set priorities in addition to executing on them.
What You'll Work On:
Supply/Demand Modeling
- Build fill-rate models for serverless inference, balancing utilization against performance and latency targets.
- Model how traffic is shared and multiplexed across models and customers to set safe capacity buffers.
Pricing Optimization
- Design pricing for low-utilization capacity, including spot and preemptible offerings, off-peak and overnight windows, and sub-scale capacity pools.
- Develop pricing policies to optimize our capacity utilization and price-yield.
- Design execute experiments and causal studies to estimate elasticity across customer segments, models, and products.
GPU Mix Optimization
- Optimize the model-hardware matching across GPU generations, vendors, clouds, and regions, accounting for workload fit, substitution, and lead times.
- Optimize our capacity across commitment types and terms, quantifying the tradeoff between cost, flexibility, and utilization risk to guide how we source capacity.