Automation × Data analysis
I build automations
and analyze data.
I build systems that take over repetitive work, then dig into the data they produce to find out why things happen, and what to do about it.
Automation
Systems that take over repetitive work: reading and routing invoices, replaying a labeled test corpus to audit the pipeline itself. Each writes a uid-keyed audit trail, so the work it does becomes data worth analyzing.
AI Invoice Intake & Approval
A production-style n8n pipeline (v4): Mistral OCR extracts invoices (including scans), business rules validate them, every run lands in an audit trail keyed by uid, and humans approve only what matters.
Data analysis
Investigations that start with a question, find the root cause in the data, and end with a clear, costed recommendation, hypotheses written down and dated before the first query.
How they connect
The two halves of my work feed each other; that loop is the whole point, not a slogan.
01 / automation
Automations create the data.
Every workflow logs an audit trail: execution id, latency, page count, route. The work it does becomes a clean, honest dataset, by design.
02 / data analysis
Analysis ends where automation begins.
Every finding resolves into a trigger, a rule, a message, and a log: the exact anatomy of a workflow. A good diagnostic doesn’t end in slides; it ends in an automation backlog.
↳ The loop, run left to right: the Audit Harness replays 150 labeled invoices through the intake pipeline and measures it row-for-row against ground truth.
↳ And right to left: the $1.67M Churn Diagnostic ends in three costed moves, each one a workflow of exactly that shape, waiting to be built.