AI & Security Engineer at Fluid Attacks · Medellín, Colombia
I build AI agent platforms, MCP servers, and secure cloud infrastructure. Currently focused on integrating LLMs into the development lifecycle through tooling that's practical, typed, and production-ready.
- MCP Servers — Lightweight implementations that expose Python functions as discoverable tools via HTTP. JSON-RPC interface with async support and type safety.
- AI Agent Platforms — Production systems using PydanticAI, LLMs, and MCP for orchestrating intelligent workflows.
- Secure APIs & Infrastructure — GraphQL APIs, ETL pipelines, and AWS cloud architecture with security built in from day one.
Python TypeScript React GraphQL AWS Terraform Docker Nix MCP PydanticAI AWS Bedrock Snowflake DynamoDB
| Project | Description |
|---|---|
| http-mcp | MCP protocol server over HTTP and STDIO built on Starlette. Type-safe tools and prompts via Pydantic, scope-based authorization, lifespan state management, and an optional auth_mcp package for OAuth 2.1 with RFC 9728 discovery, Bearer token validation, Dynamic Client Registration (RFC 7591), and WWW-Authenticate error responses. |
| pymocks | Type-safe testing utilities for mocking functions, variables, classes, and HTTP endpoints in Python. Decorator and context manager API with automatic sync/async detection and signature validation at construction time. |
Computer Science at Universidad Nacional de Colombia. Full-Stack Developer certified from Coding Dojo. I speak Spanish natively, English (B2), and French (B1).