Paul John Butad / AI Engineer & Founder

From the Philippines, with a lot of coffee.

Build it.
Check it.
Improve it.

I'm Paul. I build AI systems, evaluate what works, and use that feedback to make them better.

Still solving people's problems
with technology & design.

Selected career contributions

PioneerDev.AI / Quality

AI needs a
feedback loop.

90% fewer

Production AI chat regressions after introducing LLM-as-a-judge evaluations.

Relative regressions after evaluation

Egoist Labs / Mentor time

A whole day.
Now, half an hour.

20–30 min

One mentor. Five learners. Differentiated lesson planning and task assignment, with AI.

A full dayTime to teach.

Egoist Labs / Curriculum

Human feedback.
A better next version.

Capability Studio supports updates for each 3–6-month cohort.

PioneerDev.AI / Delivery

Less repeated work.
More room to ship.

4×

Feature deployment velocity with CI/CD and automated integration tests.

The work behind it

Currently building Egoist Labs

AI engineering · Evaluation · Learning systems

01 / Selected work

Built for real people.

From AI learning systems
to my own little coffee business.

A couple from the archive

The web projects that got me here.

Luna-Latte's Website website

Bringing live cafe experience to the comfort of your home

  • Fullstack
  • Messenger Integration
  • Customer experience
Game Antena website

Browse Top Games and Streams from popular platforms like Twitch & Youtube

  • Fullstack
  • API Integrations
  • Caching
The original portfolio illustration of a developer at a desk

Code. Coffee. Boxing.

02 / The person behind the code

Hello, I'm Paul

Self-taught. Still curious.
Still figuring things out by building.

I started with websites and interfaces. These days, I build agentic systems, evaluate AI behavior, and work on learning tools at Egoist Labs.

The thread through all of it: solving people's problems with technology and design. Outside the code? Coffee, boxing, and rap music.

AI engineering

  • Agentic systems
  • LLM evaluation
  • AI workflows

Build & ship

  • Python
  • TypeScript
  • Go
  • React
  • AWS
  • CI/CD
80% less

manual regression testing timeQuantrics · Test automation

50% lower

system costsProsple · Production system redesign

Where I've been building.

A little more context.

  • Architected an agentic AI system with serverless, event-driven pipelines for image processing.
  • Deployed an LLM-as-a-judge evaluation framework that reduced production AI chat regressions by 90%.
  • Built CI/CD pipelines and automated integration tests that increased feature deployment velocity by 4×.

03 / Notes from the process

Things I had to learn.

Written down, in case
they help someone else.