Humbert Costas

I've built security data platforms at Nestlé, including streaming ingestion for 40 TB/day of telemetry and batch ETLs processing 20 TB/day. I enabled Global SOC access to roughly 6 PB of historical security data through Azure Databricks and reduced storage from about 6 PB to 2 PB, saving over 20% of cost.

As an independent consultant and researcher, I deliver security architecture, data engineering, and AI systems for regulated sectors. I build KGCS, a cybersecurity knowledge graph, and agentdef, a portable AI agent definition specification, while working with local LLMs and agentic AI tooling.

My work combines cybersecurity operations, data engineering, machine learning, and solution architecture. I've co-founded a cybersecurity company, built SOCs and threat intelligence products, taught cybersecurity master's courses, and contributed open-source fixes and tools for MITRE and NIST-aligned standards.