Senior Member of Technical Staff, Multimodal AI

Responsibilities

• Design and develop production-grade multimodal AI systems that integrate text, vision, and speech modalities.
• Research and implement novel multimodal representation and transfer learning methods; run experiments on large-scale compute.
• Build and maintain scalable training and data pipelines for large multimodal models, including distributed training workflows.
• Optimize model inference and training performance (memory, latency, GPU utilization) for real-world deployment.
• Create robust evaluations, benchmarks, and analysis to measure capability, reliability, and safety for multimodal tasks.
• Collaborate cross-functionally with research, engineering, product, and QA teams to deliver end-to-end features and launches.

Requirements

• Strong software engineering
• Expert Python skills
• Deep learning frameworks
• Distributed training
• Multimodal experience
• Autoregressive models
• GPU optimisation
• Large-scale systems

Preferred Qualifications

• Publications in top-tier ML venues
• CUDA / GPU kernel experience
• Experience with JAX
• Production ML deployment
• Experience with vision & speech tasks
• Prior work on large-parameter models

Benefits

• Competitive salary and relocation allowance
• Health and medical insurance coverage
• Annual flight allowance / ticket
• Generous paid leave (six weeks vacation) and parental leave top-up
• Remote-flexible / hybrid working and co-working stipend
• Learning and personal enrichment budget

About the Company

Cohere builds frontier language and multimodal AI models used by developers and enterprises to power real-world products. We combine research excellence with production-grade engineering to deliver performant, reliable models. This role is operated from the UAE (Dubai-based/hybrid) and is part of a globally distributed, remote-friendly team that values diversity, inclusion, and rigorous technical work.