VP Engineering — Foundation Models

Role overview

Pantheion AI is building Pantheion-1 — the first sovereign Arabic/English frontier large language model, trained on GCC data, deployable in classified environments, and governed by Constitutional AI principles adapted for the MENA cultural and regulatory context. As VP Engineering for Foundation Models, you will lead the team responsible for training, evaluating, fine-tuning, and scaling this model from initial training runs through to production deployment for sovereign government and enterprise clients.

What you will do

• Lead the end-to-end training lifecycle for Pantheion-1 — pre-training, instruction tuning, RLHF, and Constitutional AI alignment — on 100–200 MW of classified-ready GCC compute
• Build and manage a world-class foundation model engineering team of 50–150 researchers and engineers, recruiting globally with a strong focus on Arabic AI and multilingual NLP talent
• Define Pantheion AI's model evaluation framework — Arabic language benchmarks, safety evaluations, classified-context performance testing, and GCC regulatory compliance assessments
• Oversee the development of Pantheion's specialized model fleet: domain-specific models for defense, healthcare, banking, energy, and legal verticals
• Establish the Applied Research Lab: mechanistic interpretability research, Constitutional AI governance methodology, and red-team adversarial testing operations
• Design the technical architecture for sovereign fine-tuning pipelines that allow enterprise and government clients to adapt Pantheion-1 on their own classified data without model weights leaving secure environments
• Lead open-source foundation model strategy — identifying, evaluating, and licensing base models (Llama, Mistral, Qwen) as cost-effective starting points for Pantheion's proprietary development
• Represent the foundation models team in technical reviews with sovereign government clients, investors, and regulatory bodies

Skills profile

Required qualifications

Preferred qualifications