andrewbrooks.fyi
Product sheet Read this first
Product

Andrew Brooks

ModelTechnical product manager

Nameplate

Service
15 years, enterprise software
Location
Twin Cities, Minnesota
Status
Open to senior technical and platform product roles

1.0 Description

I’m a technical product manager. Fifteen years in enterprise software: QA, engineering, then product. Today I manage a small team, set product priorities for a cloud evidence platform, and make the go or no-go call on about fifteen releases a year. On my own time, I build with AI.

Operating principle

I like problems where the product decision and the technical one are the same decision.

  1. 2.0Service historyp. 02
  2. 3.0Capabilitiesp. 03
  3. 4.0Operating principlesp. 04
  4. 5.0Contactp. 05
2.0

Service history

Every roleOldest to newest
  1. Andrew as a QA associate, inspecting for bugs
    2.1Quality Assurance Associate2011 to 2013Pivotal Vision
  2. Andrew as a systems specialist, racking a server on site
    2.2System and Application Specialist2013 to 2015Pivotal Vision
  3. Andrew as a software engineer, writing code at a keyboard
    2.3Software Engineer2015 to 2018Pivotal Vision
  4. Andrew as a product manager, placing sticky notes on a roadmap
    2.4Product Manager2018 to 2021Pivotal Vision
  5. Andrew as a product owner, ordering the backlog
    2.5Product Owner2021 to 2023Getac Video Solutions
  6. Andrew as a delivery manager, moving boxes to the new platform
    2.6Product Delivery Manager2023 to 2025Getac Video Solutions
  7. Andrew as a senior delivery manager, coaching a teammate with a GO flag
    2.7Senior Product Delivery Manager2025 to nowGetac Video Solutions
  8. Andrew as a builder, directing a row of six small robots
    2.8Builder2026On his own time
Full record
3.0

Capabilities

Capability 3.1Day job, 2023 to now

Turning migration into a product

Moving customers from our legacy single-server SaaS to a scaled, distributed microservice platform was a one-off project, run by hand each time. I turned it into a product: a repeatable path with automation and validation, and an operating model for the organization around it, covering who owns each step, how failures get analyzed, and how status is tracked. Now I lead the whole effort, across every team that touches it.

Read how it was built
Capability 3.2Day job

Helping retain a $10M+ account

Our largest domestic customer was under pressure from competitors selling features we didn’t have yet. We answered with proof, on their own system, with their leadership in the room. They renewed for five more years.

Read the field notes
Capability 3.3Day job

When the link breaks, the rule doesn’t follow

A failure in an automatic linking step meant rules reached only some of the records users expected them to cover. I contained it in three tiers and kept all 15 early adopters.

Read the field notes
On my own timePersonal, 2026

I build with AI to learn how agents behave when they run unattended, and what it takes to trust them. Finding so far: it’s mostly product and platform work. Roles, permissions, review, evals, and cost limits.

Capability 3.4Personal project, 2026

The agent engine

Six AI agents on a home server that build, test, and ship their own changes, with me as the only one who can merge. Here it is working.

Operate the engine
Capability 3.5Personal project, 2026

Mimir: one memory for every agent

The memory all my agents share, designed in a day and running since: one writer, plain text in git, and guarantees enforced in code, not by the model. Here’s what broke, and what I changed.

Read the field notes
Capability 3.6Buildathon, 2026

SeeThru, built in a buildathon

SeeThru turns every action in a sensitive system into a dot, raises the few that look wrong, and hands each one to a person to decide. Built with one teammate at the AiEdge Buildathon 2026; we reached the semifinals.

Read the field notes
4.0

Operating principles

Operating notesFrom the engine
  1. The agent that reads my email never holds the key to it.Procedure 3 →
  2. Agents write the code. Only a person can merge it.Procedure 2 →
  3. When an agent isn’t sure, it doesn’t guess. It asks, with a recommendation.Procedure 4 →
5.0

Contact

5.1 Service desk
Email [email protected]
5.2 Status

Open to technical and platform product roles, remote or hybrid from the Twin Cities.