Sergio Gómez Martín

A multi-agent AI architecture that dynamically generates and coordinates specialized LLM agents to decompose and solve complex general-purpose tasks.

• Designed a multi-agent AI architecture.
• Focused on dynamic generation and coordination of specialized LLM agents.
• Built around decomposition and solution of complex general-purpose tasks.

• LLM agents
• Multi-agent systems
• Agent orchestration

A hands-on laboratory for studying the language-model lifecycle from first principles, currently focused on implementing and training a small GPT-like model with PyTorch.

• Explores data preparation, attention mechanisms, and GPT architecture through focused notebooks.
• Builds core model components from scratch to understand their behavior and tensor transformations.
• Will expand into post-training and inference as the implementation matures.

• PyTorch
• LLM pretraining
• Transformer internals