About the Role
Our mission is to scale intelligence to serve humanity. We’re training and deploying frontier models for developers and enterprises who are building AI systems to power magical experiences like content generation, semantic search, RAG, and agents. We believe that our work is instrumental to the widespread adoption of AI. We obsess over what we build. Each one of us is responsible for contributing to increasing the capabilities of our models and the value they drive for our customers. We like to work hard and move fast to do what’s best for our customers. Cohere is a team of researchers, engineers, designers, and more, who are passionate about their craft. Each person is one of the best in the world at what they do. We believe that a diverse range of perspectives is a requirement for building great products. Join us on our mission and shape the future! Why this role? As a Performance Engineer in the Pre-Training team you will be responsible for optimizing the performance of our advanced language models and systems. Their primary focus is on improving key model training metrics, such as training throughput, ensuring high accelerator utilization. The team combines expertise in software engineering, machine learning, and low-level kernel design and development to design robust systems and enhance model performance. You will work on identifying and removing performance bottlenecks, develop cutting-edge training and profiling tools to help Cohere's mission of providing efficient and reliable language understanding and generation capabilities and drive innovation in the field of natural language processing. Note: We have offices in London, Toronto, New York and San Francisco, but we’re also remote-friendly! This team operates primarily between ET to CET time zones, so we’re seeking candidates in locations that align with these hours for effective collaboration.
Responsibilities
- Design and write high-performant and scalable software for training.
- Understand architectural modifications and design choices and their effects on training throughput and quality.
- Write low-level CUDA, triton kernels to squeeze every last bit of performance from our accelerators.
- Research, implement, and experiment with ideas on our supercompute and data infrastructure.
- Learn from and work with the best researchers in the field.
Requirements
- Extremely strong software engineering skills.
- Proficiency in Python and related ML frameworks such as JAX, Pytorch and XLA/MLIR.
- Experience writing kernels for GPUs using CUDA, triton, etc
- Experience using large-scale distributed training strategies.
- Familiarity with autoregressive sequence models, such as Transformers.
Qualifications
paper at top-tier venues (such as NeurIPS, ICML, ICLR, AIStats, MLSys, JMLR, AAAI, Nature, COLING, ACL, EMNLP).
Benefits
- An open and inclusive culture and work environment
- Work closely with a team on the cutting edge of AI research
- Weekly lunch stipend, in-office lunches & snacks
- Full health and dental benefits, including a separate budget to take care of your mental health
- 100% Parental Leave top-up for up to 6 months
- Personal enrichment benefits towards arts and culture, fitness and well-being, quality time, and workspace improvement
- Remote-flexible, offices in Toronto, New York, San Francisco, London and Paris, as well as a co-working stipend
- 6 weeks of vacation (30 working days!)