• Hours back every week
  • Fewer costly mistakes
  • Answers you can verify
  • Humans keep control
  • Leads qualified in seconds
  • Invoices checked before money moves
  • Shipped end to end, solo
  • Hours back every week
  • Fewer costly mistakes
  • Answers you can verify
  • Humans keep control
  • Leads qualified in seconds
  • Invoices checked before money moves
  • Shipped end to end, solo

Double your team’s output — without doubling the risk.

I build AI systems that clear your repetitive work — qualifying leads, checking invoices, answering document questions — and stop to ask a human the moment they’re unsure. More time for the work only you can do.

OPEN TO ROLES & PROJECTSARIEL MAGALSO · AI ENGINEER · REMOTE
  • Verdict
  • LedgerGuard
  • Provenance
  • Upskill Course
  • Cawyan Resort
  • Verdict
  • LedgerGuard
  • Provenance
  • Upskill Course
  • Cawyan Resort

The short version

The short version

What you get

Repetitive decisions handled in seconds — lead qualifying, invoice checks, document answers

What you keep

Final approval on anything with money or risk attached — the AI stops and asks

Why trust it

Every system publicly evaluated — pass rates, costs, and failures included

Work with me

AI Engineer — projects, contract, or full-time. Remote (GMT+8), solo end to end

Capabilities & Stack

Tools I ship with

The stack matters less than the result — these are just how I get you there.

AI Systems

Claude APIOpenAI APIRAG and pgvectorStructured outputsEvaluation suitesPrompt-injection defensesHuman-in-the-loop workflows

Application Engineering

PythonFastAPITypeScriptNext.js and ReactPostgreSQL and SupabaseRedis

Cloud & Delivery

DockerAWSTerraformVercelGitHub Actions

Business Integrations

HubSpot APIQuickBooks APISlack APIWebhooksn8nAWS SQS
About me

AI that ships, with receipts

You don’t need more AI demos—you need work off your plate without new risk on it. I’m Ariel Magalso, an AI engineer based in the Philippines, and I build the whole system solo: the product, the application, the evaluation harness, and the rules for when a human must decide.

Each project addresses a different failure mode: unsupported answers, unjustified business scores, and unsafe financial processing. The demos, evaluation results, architecture, and source code are publicly inspectable.

Contact

Hiring an AI engineer?Let's talk.

If there’s a workflow eating your team’s week, I can probably build the system that takes it over — safely. Open to projects, contract work, and full-time AI engineer roles; everything above is live and inspectable if you want to check the receipts first.

Have a role, a project, or just want to talk shop? Drop a message — I’ll get back to you.

USUALLY REPLIES WITHIN 24 HOURS · GMT+8 · REMOTE-READY