AI Researcher – Multilingual Data

About the role

Join a cutting‑edge research environment in Saudi Arabia where multilingual AI innovation meets real‑world impact. As an AI Researcher focused on multilingual data, you will lead the creation of high‑quality datasets and research strategies that power next‑generation language models across diverse languages and domains.

Key responsibilities

• Design and conduct research on multilingual datasets, including collection, filtering, deduplication, quality assessment, and optimization.
• Develop innovative strategies for low‑resource and long‑tail languages using advanced sampling, data augmentation, and curriculum learning.
• Improve multilingual large language models by enhancing cross‑lingual transfer, alignment, robustness, and representation learning.
• Build, maintain, and refine multilingual evaluation benchmarks to measure model quality across languages.
• Collaborate with machine learning engineers and researchers to influence training pipelines, model architectures, and production deployment.
• Publish research findings at leading AI and NLP conferences and contribute to open‑source initiatives when appropriate.

Required profile

• Advanced background in Natural Language Processing, Machine Learning, or a closely related field.
• Proven research experience in multilingual or cross‑lingual language modeling with publications at top conferences or journals (e.g., ACL, EMNLP, NeurIPS, ICML, ICLR).
• Hands‑on experience with large‑scale multilingual text datasets and modern ML workflows.
• Ability to work independently, manage research initiatives, and deliver high‑quality results in a fast‑moving environment.

Required skills

• Python programming.
• Deep learning frameworks such as PyTorch or JAX.

What we offer

• Opportunities to publish influential research and shape the future of multilingual AI.
• Collaboration with highly skilled experts in a fast‑paced, innovation‑driven setting.
• Potential involvement in open‑source projects and industry‑leading benchmarks.