Director of AI Research - Foundation Models & LLM Pretraining

Build the Models. Push the Boundaries. Shape the Future of AI.

Location: Remote | United States

Employment: Full-Time | Research Leadership

Please read the information in this job post thoroughly to understand exactly what is expected of potential candidates.

The Opportunity

We're partnering with an industry-leading AI technology company pushing the boundaries of foundation models, generative AI, and intelligent systems deployed at global scale.

We're seeking an exceptional AI research leader who has personally architected, built, and pre-trained large language models (LLMs) or foundation models from scratch.

This isn't about fine-tuning existing models or building applications on top of someone else's AI. It's about building the foundational intelligence itself.

You'll tackle some of the most challenging problems in modern AI, influence critical architectural decisions, and help define the next generation of intelligent systems.

What You'll Do

• Build From Scratch: Architect, develop, and pre-train large-scale language and multimodal foundation models.
• Push Architectural Boundaries: Advance transformer architectures, attention mechanisms, mixture-of-experts (MoE), and emerging model designs.
• Solve Complex Challenges: Tackle training divergence, convergence, scaling laws, and optimization across large-scale GPU clusters.
• Advance AI Performance: Improve training efficiency, model intelligence, generalization, and inference performance.
• Lead Innovation: Drive research priorities, influence technical strategy, and shape the future of production-scale generative AI systems.

What You'll Bring

• Essential: Proven, hands-on experience pre-training large-scale LLMs or foundation models from scratch — not simply fine-tuning existing models.
• Deep expertise in transformers, modern deep learning, model architecture, and optimization.
• Strong knowledge of RoPE, GQA, MoE, attention mechanisms, and scaling laws.
• Experience with distributed GPU training, FSDP, DeepSpeed ZeRO, tensor parallelism, and pipeline parallelism.
• Hands-on experience with BF16/FP8, gradient checkpointing, training stability, and large-scale GPU infrastructure.
• Demonstrated ability to lead complex research initiatives and influence architectural decisions.

Bonus: Multimodal foundation models, RLHF, DPO, GRPO, or significant contributions to advanced AI research.

Why This Opportunity?

You're not joining to build on top of existing foundation models. You're joining to help create the next generation.

This is an opportunity to combine groundbreaking research, technical ownership, and real-world impact — solving problems that will shape how intelligent systems are built and deployed worldwide.

If you've built foundation models from scratch and want to push the boundaries of what's possible in AI, we'd love to connect. xhyhwjd

Apply today to explore this confidential research leadership opportunity.