Ezequiel Gomez

- Defined and led Meta’s 2025 IAM strategy for AI agents, establishing the company-wide access model adopted by internal and third-party agents across Meta’s infrastructure.

- Architected Meta’s identity attributes primitive for agent authentication, allowing backend services to distinguish direct human actions from agent-mediated actions using the same human identity; adopted across Meta’s infrastructure for audibility, access control, and agent-specific policy enforcement.

- Removed memory bottlenecks in Meta’s Authorization backend (handling 40B+ QPS) which capped the system at ~2.5M unique access-control policies. Eliminating OOMs as a recurring problem, enabling continued policy growth, and reducing memory utilization by ~45%.

- Built Meta’s Identity Management Service, the authoritative registry for internal human and non-human identities and identity attributes, consolidating 15+ fragmented datastores and powering 1M+ roles and dynamic access policies across Meta’s infrastructure.

- Delivered new Mandatory Access Control product, enabling central security governance teams to enforce population-based guardrails over sensitive permissions.

- Served as long-term technical owner for CATs, Meta’s core authentication token. Used to authenticate services, workloads, agents, external users, and employees. Drove its adoption and scaled the system to 3B+ token creation and verification events/sec and 99.999% availability.

- Backend service done in C++. Client-side library support in C++, Hack, Python, and Java.

- Implemented ElectrumX client node in C#, allowing the trading application to consume cryptographically verifiable information from Bitcoin, Litecoin or Ethereum without having to download ~150GB worth of on-chain transaction information.

- Developed a prototype of a cryptographic scheme to enable anonymous on-chain transactions in C++ using the Libsnark and Sodium cryptographic libraries.