Fireworks AI
MTS, Research Engineer
New York, NY; San Mateo, CAFrom $400kmidAdded today
About this role
Join Fireworks AI as a Research Engineer to bridge cutting-edge ML research and large-scale distributed training infrastructure. You'll conduct open-ended research on novel architectures and optimization techniques while building and optimizing the systems needed to train massive models efficiently on GPU clusters.
What you'll do
- Conduct open-ended research on model architectures, training objectives, and optimization techniques
- Reproduce state-of-the-art results from ML papers and identify improvements at scale
- Design and maintain high-performance distributed training systems for large GPU clusters
- Optimize training loops, data pipelines, and communication overhead
- Translate mathematical concepts into efficient, production-grade code
- Collaborate with research scientists to co-design hardware-aware experiments
What they're looking for
- Python, C++, or Rust programming
- PyTorch, JAX, or TensorFlow expertise
- Distributed systems and parallel computing (CUDA, NCCL, MPI)
- Deep learning algorithm implementation
- Linear algebra, calculus, probability, and statistics
- GPU programming (CUDA/Triton preferred)
- Large Language Model training experience
- Inference optimization and OSS engines (SGLang, vLLM)
Benefits
- Meaningful equity in a Series C AI infrastructure company
- Competitive base salary ($250K–$400K USD)
- Comprehensive benefits package
- Work on cutting-edge AI infrastructure problems
- Collaborative environment with world-class researchers and engineers
- No bureaucracy, ownership-driven culture
Likely interview questions
- Describe a time you implemented a complex deep learning algorithm from scratch—what were the biggest challenges and how did you overcome them?
- How have you optimized training loops or data pipelines for distributed GPU clusters, and what metrics did you focus on?
Opens the official application on the employer’s site. No login required.