AI Engineer Intern – LLM Data

About the role

MBZUAI is a research institute for foundation models, and it is recruiting an AI Engineer Intern specializing in LLM data. The intern works with the team to build and improve high-quality training data for foundation models across pre-training, mid-training and post-training, gaining hands-on experience with large-scale data processing, LLM-based data synthesis, data quality evaluation and model experimentation.

Key responsibilities

• Build, curate and improve datasets for LLM pre-training, mid-training and post-training, including urgent requests that need a fast turnaround.
• Develop and improve data processing workflows for extraction, cleaning, filtering, deduplication, transformation and quality control.
• Use LLMs to generate, refine, filter and evaluate synthetic data.
• Analyze data quality and identify issues such as duplication, low-quality samples, distribution gaps and coverage limitations.
• Run experiments to see how different data sources and processing strategies affect model performance.
• Support model inference, evaluation and training experiments when needed to validate data quality.
• Research new datasets, data processing techniques and LLM data methodologies.
• Collaborate closely with researchers and engineers on fast-moving foundation model projects.

Required profile

• Currently pursuing a Bachelor's, Master's or PhD in Computer Science, AI, Machine Learning, Data Science or a related field.
• Hands-on experience with LLMs through research, coursework, projects, internships or open-source work.
• Comfortable working with datasets and performing data processing and analysis.
• Strong problem-solving skills and ability to work independently in a fast-paced research environment.

Required skills

• Strong programming in Python.
• Good understanding of machine learning, deep learning, NLP or large language models.
• Preferred: PyTorch, Hugging Face, vLLM or similar frameworks; LLM data synthesis, fine-tuning, model evaluation and prompt-based generation; familiarity with LLM training pipelines; experience with large-scale datasets or distributed data processing; relevant research or open-source contributions.