Skip to content

I put AI inside production systems. Then I audit what it writes.

Since 2011 I’ve been fixing and modernizing systems businesses run on. Now with models, agents, and MCP servers in the mix, and a review of the code they produce before it ships.

Main tools: Java · Spring Boot · .NET · Node · Python · PostgreSQL · REST APIs · models · agents · MCP

See the work Email me
Portrait pending
Since
2011, continuously
Tech lead
Since 2017
Countries
Venezuela, Chile, Argentina, Ecuador, Mexico, and Spain
Languages
Native Spanish, English B1

What I fix

  1. Legacy nobody wants to touch

    Java 8 and JSP apps, old .NET, modules only one person understands. I read, I document, and only then do I change anything.

  2. AI shipped without review

    Copilots writing code nobody reads, prompts with customer data, agents with more permissions than they need. I review what goes in, what comes out, and what gets logged.

  3. Queries that collapse under history

    Reports that take minutes on huge tables. I measure, split, index, and measure again.

  4. Friday deploys at 11 p.m.

    No CI/CD, no rollback, no staging environment. I build the path so a deploy stops being an event.

  5. Systems that don’t talk to each other

    Third-party APIs, the same data in three databases, manual integrations. I define the contract and write it down.

  6. Teams without shared standards

    Everyone deploys differently and nobody reviews code. I set up reviews, branching, and rules the team can keep.

Applied AI

Models inside the workflow

I integrate models from different providers via API into internal tools the team opens every day.

Limit: if the task doesn’t repeat, it doesn’t justify a model.

Agents with scoped tools

Clear permissions, defined scope, and a log of what they do.

Risk: an agent without limits is an incident waiting its turn.

MCP servers

I connect models to in-house tools and data using MCP (Model Context Protocol), with the minimum access the task needs.

Limit: what the agent can’t read, it can’t leak.

Auditing generated code

I read what the model writes before it ships: input validation, contracts, error handling, dependencies, and tests.

Risk: fast isn’t the same as correct.

Knowing when not to use it

A model won’t fix a missing index, a badly designed data model, or a process nobody defined.

Limit: when the answer is code and not AI, I say so.

Nobody read what the model wrote. The deploy shipped anyway. Where I draw the line with AI

How I work

  1. Business first

    I don’t touch a line until I know what the system does and who uses it.

  2. Small, verifiable changes

    One thing at a time, with tests where possible. If the code came from a model, someone read it before it merged.

  3. Better than I found it

    Documented code, recorded decisions, and someone on the team who understands what I did.

Selected work

Chile · –

B2B telecom channel

Context
Legacy applications in the B2B channel running critical sales and billing processes.
What I did
I optimized Java 8 and JSP components, integrated external APIs, and supported the development of new applications.

Stack: Java 8 · JSP · REST APIs

Argentina ·

Procurement, tenders, and auctions

Context
A transactional platform with purchasing, tendering, and user management processes.
What I did
Spring Boot backend services, frontend components, and REST APIs over PostgreSQL connecting the modules.

Stack: Spring Boot · React · PostgreSQL

Overtime tracking

Context
Internal logging and tracking of overtime hours.
What I did
.NET REST APIs, SQL Server integration, and deploys to Azure.

Stack: .NET · Angular · SQL Server · Azure

See all cases

Stack

Every day

  • Java
  • C#/.NET
  • PHP
  • JavaScript and TypeScript
  • SQL
  • Git

Applied AI, daily

  • Models via API from different providers
  • Agents with tools
  • MCP servers
  • Auditing generated code
  • Criteria for when to use AI and when not to

When the project calls for it

  • Spring Boot
  • Node.js
  • Python
  • Laravel
  • Angular
  • Vue
  • React
  • REST APIs
  • GraphQL
  • MongoDB
  • Redis
  • Docker
  • CI/CD

Enough not to break it

  • Azure
  • AWS
  • Kubernetes
  • Terraform

Recent writing

Got a project in mind?

Tell me what broke. There’s a person on the other side, not a bot.

guajilodev@gmail.com Download CV