Cohere

Member of Technical Staff, Training Performance Engineer

London, England, United Kingdom

Not SpecifiedCompensation
Mid-level (3 to 4 years), Senior (5 to 8 years)Experience Level
Full TimeJob Type
UnknownVisa
AI & Machine Learning, Data & Analytics, Enterprise SoftwareIndustries

Requirements

Candidates should have extremely strong software engineering skills and proficiency in Python along with related ML frameworks such as JAX, Pytorch, and XLA/MLIR. Experience writing kernels for GPUs using CUDA and Triton is required, as well as experience with large-scale distributed training strategies. Familiarity with autoregressive sequence models, such as Transformers, is also necessary.

Responsibilities

As a Performance Engineer, you will design and write high-performant and scalable software for training. You will understand architectural modifications and their effects on training throughput and quality. The role involves writing low-level CUDA and Triton kernels to optimize performance, researching and implementing ideas on supercompute and data infrastructure, and learning from and collaborating with top researchers in the field.

Skills

Python
JAX
Pytorch
XLA
MLIR
CUDA
Triton
Distributed Training
Transformers
Software Engineering
Kernel Design

Cohere

Provides NLP tools and LLMs via API

About Cohere

Cohere provides advanced Natural Language Processing (NLP) tools and Large Language Models (LLMs) through a user-friendly API. Their services cater to a wide range of clients, including businesses that want to improve their content generation, summarization, and search functions. Cohere's business model focuses on offering scalable and affordable generative AI tools, generating revenue by granting API access to pre-trained models that can handle tasks like text classification, sentiment analysis, and semantic search in multiple languages. The platform is customizable, enabling businesses to create smarter and faster solutions. With multilingual support, Cohere effectively addresses language barriers, making it suitable for international use.

Key Metrics

Toronto, CanadaHeadquarters
2019Year Founded
$914.4MTotal Funding
SERIES_DCompany Stage
AI & Machine LearningIndustries
501-1,000Employees

Risks

Competitors like Google and Microsoft may overshadow Cohere with seamless enterprise system integration.
Reliance on Nvidia chips poses risks if supply chain issues arise or strategic focus shifts.
High cost of AI data center could strain financial resources if government funding is delayed.

Differentiation

Cohere's North platform outperforms Microsoft Copilot and Google Vertex AI in enterprise functions.
Rerank 3.5 model processes queries in over 100 languages, enhancing multilingual search capabilities.
Command R7B model excels in RAG, math, and coding, outperforming competitors like Google's Gemma.

Upsides

Cohere's AI data center project positions it as a key player in Canadian AI.
North platform offers secure AI deployment for regulated industries, enhancing privacy-focused enterprise solutions.
Cohere's multilingual support breaks language barriers, expanding its global market reach.

Land your dream remote job 3x faster with AI