Member of Technical Staff, Data Analysis and Evaluation

Responsibilities

• Design and oversee data collection tasks, including managing and supporting human annotators and ensuring annotation quality within UAE-based projects.
• Develop and apply statistical methods to evaluate dataset quality, annotation reliability, and experiment validity.
• Analyse model robustness and generalisability across diverse use cases, datasets, and languages relevant to the UAE region.
• Collaborate with researchers, engineers, and data teams to improve dataset curation and model performance.
• Train and fine-tune large language models (LLMs) on distributed training infrastructure and run evaluation experiments.
• Translate experimental findings into actionable recommendations to improve production systems and data pipelines.

Requirements

• Software engineering
• Statistical analysis
• Experimental design
• LLM training
• Distributed training
• Python programming
• PyTorch or TensorFlow
• Data annotation
• Communication skills
• Research papers

Preferred Qualifications

• Publications in top-tier venues (NeurIPS, ICML, ICLR, etc.)
• Experience with JAX or ML infra (training clusters, orchestration)
• Background in evaluating bias, fairness, and dataset quality
• Experience running large-scale human annotation programs
• Proven track record improving model generalisability and robustness

Benefits

• Health Insurance
• Medical Insurance
• Annual Leave
• Paid Leave
• Transportation
• Visa support

About the Company

Cohere builds frontier language models and tools that power semantic search, generation, RAG, and agent experiences. We focus on research-driven product development and ship impactful ML systems. We welcome applicants based in the UAE (Dubai, Abu Dhabi, and other emirates) — this role supports UAE-based candidates and can be performed remotely within the country or hybrid in-office when available. Cohere values diverse perspectives and inclusive collaboration across engineering, research, and data teams to scale intelligence for humanity.