Research Engineer

Responsibilities

• Conduct open-ended research on new model architectures, training objectives, and optimization techniques
• Reproduce and extend state-of-the-art results from recent machine learning papers, identifying bottlenecks and scaling methods
• Design, implement, and maintain high-performance distributed training systems across large GPU clusters
• Translate abstract mathematical concepts and research ideas into robust, efficient code
• Collaborate with Research Scientists to unblock experiments through tooling, optimization, and hardware-aware co-design

Requirements

• Strong programming skills in Python, C++, or Rust with a commitment to clean, maintainable code
• Deep practical knowledge of machine learning frameworks such as PyTorch, JAX, or TensorFlow
• Experience working with large distributed systems and parallel computing (e.g., CUDA, NCCL, MPI)
• Strong foundation in linear algebra, calculus, probability, and statistics
• Proven track record of implementing complex deep learning algorithms from scratch

Preferred Qualifications

• Master’s or PhD in Computer Science, Machine Learning, Physics, Mathematics, or related field (or equivalent industry experience)
• Experience with low-level GPU programming (CUDA/Triton) or hardware co-design
• Familiarity with challenges of training Large Language Models (LLMs)
• Familiarity with inference and OSS inference engines such as SGLang and vLLM

About the Company

Fireworks AI builds generative AI infrastructure, delivering high-quality models with fast, scalable inference. Independently benchmarked as a leader in LLM inference speed, the company is a Series C startup valued at $4 billion and backed by Benchmark, Sequoia, Lightspeed, Index, and Evantic. Founded by veterans of Meta PyTorch and Google Vertex AI.