MTS, Research Engineer

Fireworks; San Mateo (CA); On-site; USD 140,000 - 200,000; Full time

Job summary

Fireworks is seeking a Research Engineer to join the team at the intersection of model research and training infrastructure. You will design novel architectures, reproduce state-of-the-art results, and build the distributed systems needed to scale experiments on large GPU clusters.

You’ll bridge science and engineering by translating theoretical ideas into robust, performant code, collaborating with researchers to push the envelope in enterprise AI infrastructure.

Qualifications

• Strong programming skills in Python, C++, or Rust.
• Deep practical knowledge of ML frameworks (PyTorch, JAX, or TensorFlow).
• Experience with large distributed systems and parallel computing (CUDA, NCCL, MPI).
• A strong foundation in linear algebra, calculus, probability, and statistics.
• A proven track record of implementing complex deep learning algorithms from scratch.

Responsibilities

• Conduct Open-Ended Research: Explore new model architectures, training objectives, and optimization techniques. Formulate hypotheses, design experiments, and iterate quickly based on empirical results.
• Reproduce and Extend State-of-the-Art: Implement and reproduce results from recent machine learning papers. Identify bottlenecks, propose improvements, and scale these methods to larger datasets and models.
• Build and Scale Training Infrastructure: Design, implement, and maintain high-performance, distributed machine learning systems. Optimize training loops, data loaders, and communication overhead across large GPU clusters.
• Bridge Science and Engineering: Translate abstract mathematical concepts and research ideas into robust, bug-free, and efficient code.
• Collaborate Cross-Functionally: Work closely with Research Scientists to unblock their experiments by providing tooling, optimizing code, and co-designing experiments that are hardware-aware.

Skills

Python

C++

Rust

Education

Master’s or PhD in CS/ML/Physics/Math

Tools

PyTorch

JAX

TensorFlow

Job description

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.

About the Role

We are looking for a Research Engineer to join our team, operating at the critical intersection of model research and training infrastructure.

In this role, your time will be split between tackling open-ended research problems—such as designing novel architectures and improving algorithmic efficiency — and building the distributed training systems required to make those research breakthroughs a reality. You won’t just be handed a paper to implement; you will be expected to reproduce state-of-the-art results from the literature, identify their limitations, and build the infrastructure needed to push beyond them.

The most significant advances in deep learning require massive scale. We need engineers who are as comfortable reasoning about gradient descent and loss landscapes as they are about distributed systems, GPU cluster utilization, and data pipelines.

What You’ll Do

• Conduct Open-Ended Research: Explore new model architectures, training objectives, and optimization techniques. Formulate hypotheses, design experiments, and iterate quickly based on empirical results.

• Reproduce and Extend State-of-the-Art: Implement and reproduce results from recent machine learning papers. Identify bottlenecks, propose improvements, and scale these methods to larger datasets and models.

• Build and Scale Training Infrastructure: Design, implement, and maintain high-performance, distributed machine learning systems. Optimize training loops, data loaders, and communication overhead across large GPU clusters.

• Bridge Science and Engineering: Translate abstract mathematical concepts and research ideas into robust, bug-free, and efficient code.

• Collaborate Cross-Functionally: Work closely with Research Scientists to unblock their experiments by providing tooling, optimizing code, and co-designing experiments that are hardware-aware.

We Expect You To Have:

• Strong programming skills (Python, C++, or Rust) and a commitment to writing clean, maintainable code.

• Deep practical knowledge of machine learning frameworks (PyTorch, JAX, or TensorFlow).

• Experience working with large distributed systems and parallel computing (e.g., CUDA, NCCL, MPI).

• A strong foundation in linear algebra, calculus, probability, and statistics.

• A proven track record of implementing complex deep learning algorithms from scratch.

Nice to Have:

• A Master’s or PhD in Computer Science, Machine Learning, Physics, Mathematics, or a related field (or equivalent industry experience).

• Experience with low-level GPU programming (CUDA/Triton) or hardware co-design.

• Familiarity with the challenges of training Large Language Models (LLMs)

• Familiarity with the challenges of inference, and OSS inference engines such as SGLang and vLLM

Why Fireworks?

• Solve Hard Problems: Tackle challenges at the forefront of AI infrastructure, from low-latency inference to scalable model serving.

• Build What’s Next: Work with bleeding-edge technology that impacts how businesses and developers harness AI globally.

• Ownership & Impact: Join a fast-growing, passionate team where your work directly shapes the future of AI—no bureaucracy, just results.

• Learn from the Best: Collaborate with world-class engineers and AI researchers who thrive on curiosity and innovation.

Fireworks AI is an equal-opportunity employer. We celebrate diversity and are committed to creating an inclusive environment for all innovators.