Skip to content
View maxwellwilber-cpu's full-sized avatar

Block or report maxwellwilber-cpu

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
maxwellwilber-cpu/README.md

Maxwell Wilber

I build AI automations and agents for businesses.

Currently an AI implementation specialist at a Seattle AI startup, where I turn business workflows into agentic systems that run on their own. Before that I spent a year building AI systems for small business owners through an independent consulting practice, and five years running operations for a $2.5M home staging company.

Business degree, self-taught on the technical side. I care most about the part people skip: making sure the output is right, and that the system does not quietly break two months later.

What I work on

  • Agentic workflows and AI agents triggered by real business events
  • Data pipelines that turn messy multi-source business data into something decisions can be made from
  • Process automation across CRMs, email, databases and APIs
  • Validation and evaluation: testing AI output instead of trusting it

Stack: Python (pandas, pytest), SQL, Claude API and Claude Code, Postgres, REST/webhook integration, Zapier / Make / n8n, Vellum

Selected work

Project What it is
hubspot-audit Point it at a HubSpot portal, get back what is broken in the CRM with the rule behind every number and the record IDs to fix. It cannot write to your CRM, and that is enforced by ten tests rather than promised. 215 tests.
client-data-cleaner Turns several messy customer exports into one clean master list. 100% precision and 94.6% recall against generated data with known answers, zero wrong merges, with the accuracy harness in the repo so anyone can re-run it.
checkpoint Validates AI output against rules you declare. Catches invented numbers, citations pointing at nothing, and placeholder text that schema validation misses. 92 tests.
evs 73 checks with 43 pytest tests, validating the nine tables an AI pipeline produces before anyone acts on them. Catches all 12 defects planted in the sample data, plus 2 nobody planted.
fas-case-study Architecture case study of a 27-node AI financial analysis pipeline: validation design, fail-closed behavior, real client results.
ai-implementations-portfolio 10 production AI systems and 9 strategy assets built for two businesses, with detail pages on the five largest.

Seattle, WA · LinkedIn

Pinned Loading

  1. hubspot-audit hubspot-audit Public

    Read-only CRM hygiene audit for HubSpot. 25 rules, every finding traced to the rule that produced it.

    Python

  2. checkpoint checkpoint Public

    Validate AI output against rules you declare. Catches invented numbers, bad citations, and placeholder text that schema validation misses.

    Python

  3. client-data-cleaner client-data-cleaner Public

    Turn several messy customer exports into one clean master list. With every merge logged and the matching accuracy measured.

    Python

  4. ai-implementations-portfolio ai-implementations-portfolio Public

    10 production AI systems and 9 strategy assets built for two businesses, with detail pages on the five largest.

  5. fas-case-study fas-case-study Public

    27-node AI financial analysis pipeline — case study on validation architecture, fail-closed design, and real-world results

  6. evs evs Public

    73-check Python validation framework for AI-generated financial analysis output

    Python