Principal AI/ML Engineer

Charterhouse is working with a specialist AI consultancy based in the UAE that partners with enterprise clients to deliver and implement cutting-edge Generative AI solutions.

As part of continued growth, they are looking to expand their delivery team with the addition of a Principal AI/ML Engineer.

The Principal AI/ML Engineer will take ownership of architecting and delivering intelligent AI solutions and workflows, leveraging Python, PyTorch, and modern generative AI frameworks to build production-grade solutions.

The role involves fine-tuning models using approaches such as LoRA, qLoRA, and full fine-tuning, optimising prompts, embeddings, and retrieval strategies, and building and deploying retrieval-augmented generation (RAG) solutions.

The Principal Engineer will also oversee and contribute to developing custom pipelines for document ingestion, chunking, embedding generation, and retrieval, working with vector databases and RAG evaluation frameworks.

Beyond model development and deployment, the role also requires a strong understanding of AI performance and deep learning optimisation.

The Principal Engineer will profile and benchmark AI workloads across hardware platforms, experiment with model compression techniques including quantization, pruning, and knowledge distillation, and bring a working knowledge of how deep learning frameworks such as PyTorch and TensorFlow behave at a lower level.

Given the complexity of enterprise AI deployments, hands-on experience deploying and optimising AI workloads across GPU, TPU, or NPU infrastructure is strongly preferred.

The ideal candidate will bring strong software engineering foundations including testing, debugging, documentation, and version control, alongside experience with MLOps tooling such as ClearML, Kubeflow, or MLflow and data analysis libraries including Pandas, NumPy, and SQL.

A strong foundation in large language models (LLMs) is essential for this role, with the ideal candidate bringing hands-on experience across multiple LLMs and a solid understanding of how different models can be applied and optimised for domain-specific use cases.