Applied AI · Workflow systems · Product engineering

Practical AI
Systems for
Messy Manual
Workflows.

I design and build AI-powered products, knowledge systems and automation that make complex work simpler.

An intricate spherical network of connected nodes and paths

Systems shipped

End-to-end

From messy input to a product people can operate on a normal Tuesday.

Focus

AI + workflow

Products that organise dense information and make the next decision obvious.

Stack depth

Full build

Product design, AI-assisted flows, Next.js / Astro, Supabase, CMS architecture.

Michael Low portrait
Michael Low · São Paulo

About

I make complicated work easier to operate.

I work between product design, engineering and operations — usually where the problem is still ambiguous and the workflow has not been properly defined yet.

I build AI workflows and systems for people who have too much input and not enough structure. I like messy problems — half-broken processes, too many tabs, scattered notes, and the sense there must be a better way. That is usually where I come in.

Based in São Paulo · Working globally Open to AI product & systems roles Full about page

Capabilities

What I bring to an AI team.

Not slide decks — shipped systems. Comfortable across product, AI workflow design and the engineering needed to make them real.

Product

  • Product design
  • Information architecture
  • Decision-centred UX

AI systems

  • AI-assisted workflows
  • Knowledge systems
  • Retrieval & notes ops

Engineering

  • Next.js & Astro
  • Supabase
  • Automation & CMS

How I work

Refinement is the work.

I rarely start with a clean specification. I start with scattered information, repeated friction and a problem that has not been properly framed yet.

  1. 01

    Observe

    Map the real workflow — including the awkward workarounds people have learned to ignore.

  2. 02

    Simplify

    Cut the problem down to the decisions, information and hand-offs that actually matter.

  3. 03

    Build

    Ship a usable product or operating layer people can run without constant explanation.

  4. 04

    Iterate

    Watch it in use, find the friction, and refine until the structure feels inevitable.

Next step

Looking for someone who can turn messy AI problems into working systems?

Open to AI product, workflow and systems roles — remote or hybrid — where the work is shipping useful software, not theatre.