AI & Security Engineer · Medellín, Colombia

Yeison Liscano

I build secure AI agent platforms and the Python tooling behind them.

My work spans MCP servers, PydanticAI agents, GraphQL APIs, React frontends and AWS infrastructure, with security designed in from the start.

See projectsRead the blog

Projects

Selected work

Open-source Python library

http-mcp

A lightweight Model Context Protocol server for Starlette and FastAPI apps. It exposes Python functions as MCP tools and prompts over HTTP or STDIO.

  • Stateless: no sessions, so it scales horizontally behind a load balancer
  • Pydantic models generate each tool's schema and validate every call
  • Scope-based authorization per tool, plus OAuth 2.1 support
  • Python
  • Pydantic
  • Starlette
  • JSON-RPC
GitHub PyPI How it was built

Full-stack e-commerce · live

Atal Distributions

A cosmetics store with a product advisor and an admin panel for inventory, supplier invoices, restocking, sales and accounting.

  • Rust (axum) API on AWS Lambda, with a single DynamoDB table
  • Paying or cancelling an order is one conditional transaction, so a decision never applies twice
  • The accounting close covers the income statement, the IVA position and CSV exports
  • Astro
  • Rust
  • AWS Lambda
  • DynamoDB
  • CloudFront
Visit site Private repository

Full-stack e-commerce · live

Haejy

A T-shirt store with a customer storefront, a size finder, and an admin panel for orders, inventory, supplier invoices and sales.

  • Each product is a size × color matrix, not a flat list of variants
  • Orders reserve stock atomically while payment is coordinated
  • The size finder computes a fit per garment from body measurements
  • Astro
  • Rust
  • AWS Lambda
  • DynamoDB
  • CloudFront
Visit site Private repository

Open-source Python testing library

pymocks

Decorators and context managers for mocking functions, variables and HTTP endpoints in tests.

  • Detects sync and async test functions automatically
  • Every utility works as a decorator or a context manager
  • Python
  • pytest
  • asyncio
GitHub PyPI

Experience

AI & Security Engineer

Fluid Attacks

  • Architected a production AI agent platform using MCP, LLMs and PydanticAI.
  • Built the remote MCP server that lets customers query their platform data from LLM clients, with scoped tool access and OAuth 2.1 authorization.
  • Designed http-mcp, the stateless MCP library behind that server, and published it as open source.
  • Built and maintained GraphQL APIs, React frontends, ETL pipelines and AWS cloud infrastructure.
2023 – Present

Education

Computer Science

National University of Colombia — Medellin

2024 – Present

Full-Stack Developer

Coding Dojo — Virtual

2022 – 2023

Skills

PythonTypeScriptReactGraphQLAWSTerraformDockerNixMCPPydanticAIAWS BedrockSnowflakeDynamoDBAstro

Contact

Let's connect

Open to interesting conversations, collaborations, and new opportunities.

LinkedInGitHub