A guided look
Start with what
matters to you.
There is a fair amount here. Choose a perspective, and I’ll connect three examples to the kind of work you have in mind.
AI operations · Three projects
From a practical need to a workflow worth relying on.
I start by understanding the work, then decide what the tool needs to do and how its results can be checked. These three examples show that approach at different scales.
- 01
Choose a proportionate tool
PF AI
I built and personally used a reference assistant during my work at Planet Fitness, choosing NotebookLM for source-grounded answers and checkable citations.
- The work
- A focused way to consult club training and operating materials during day-to-day work.
- Why it matters here
- The tool choice followed a defined information need. Making an answer easy to check was part of making it useful.
Scope & statusPersonal workplace use; independent staff adoption has not been established.
- 02
Review what the system produces
Fairfield County’s Community Foundation (FCCF)
I helped define AI-assisted research, co-produce civic-landscape data and develop methodology. I questioned unsupported classifications and requested corrections and verification.
- The work
- Research workbooks, source-tracking fields, methodology notes and a feasibility proposal.
- Why it matters here
- Useful operations require a way to distinguish a supported result from a convincing answer that still needs attention.
Scope & statusCompleted research engagements; the proposal is not evidence of an implemented program.
- 03
Coordinate development responsibly
Aaron’s Autonomous Alpha (AAA)
I designed an autonomous trading system and direct AI-assisted implementation and review, defining requirements, acceptance criteria and reusable agent skills.
- The work
- System design, source implementation, evaluation requirements and authored development tools.
- Why it matters here
- The work connects task ownership, tool choice and evidence required before relying on a consequential system.
Scope & statusActive development focused on PAPER/testnet validation. Live trading is disabled; continuous autonomous trading and economic edge are not established.