Things I've built, drawn as the systems they are, with what I learned building each one.
B-01 · fleet-pm
I’ve used it to direct AI-assisted development of more than 10 applications, in fitness, research, education, accounting, legal and compliance, cybersecurity, and marketing.
B-02 · Alloybase
OPTIMADE gives many materials databases the same query language, so one search can cover several of them. The hard part is everything after the search: keeping track of which database each row came from, versioning a dataset so a paper can cite it, and getting it into pandas or Jupyter. When a database is slow or down, Alloybase shows the results it has and says which sources are missing.
B-03 · Member RAG chatbot
10,000members the chatbot serves
I took over my company’s chatbot when our full-stack developer left and started over in Langflow and custom Python. We’ve since shipped tailored versions for two product launches.
The main thing I learned is where the model should and should not sit. My first version was an all-in-one agent that made decisions and ran the retrieval flows itself, and it burned through tokens and gave a mediocre experience. Now a structured-output model identifies the intent, code runs the retrieval that intent needs, and every AI feature has a defined failure mode instead of a guess.
Prompts are versioned and tested against a fixed bank of common member questions, and I think that maintenance process matters as much as the first build.
B-04 · Adaptive Athlete
Coaching runs through a multi-agent language-model pipeline. Prompts ship as versioned prompt packs, and A/B tests between versions are stored in the database, so two versions of a prompt can be compared directly. It has subscription billing and runs on DigitalOcean.
B-05 · Launch and content automation
80 → 1person-hours to set up our most complex launch
BEFORE · about 80 person-hours
AFTER · about 1 hour
Creating every workflow and task for our most complex product launch went from about 80 person-hours to about one hour over three iterations (a workflow redesign, a helper script, then a Google Apps Script orchestrator).
~600 → ~7.5person-hours a year of launch setup, about 590 hours of average projected yearly savings
BEFORE · about 600 person-hours a year
15 rows = 15 launches · 40 squares = 40 person-hours each
AFTER · about 7.5 hours
Projected from an average year of about 15 launches at about 40 person-hours each, at the same reduction as the most complex launch.
100+hours saved drafting a docuseries
each dot is one hour handed back to the team
Local AI pipelines drafted the scripts for a 2025 docuseries and saved the team over 100 hours.
Both are internal tools I built at my day job, so there is nothing public to link. Launch setup went through three iterations: a workflow redesign around project phases, a helper script, then a full Google Apps Script orchestrator.