I make messy work feel easier. I'm drawn to the messy middle — the place where ideas, people, tools, and unfinished processes all meet.
My path has moved through training, business ownership, project work, operations, and now AI systems. At every stop, I found myself doing the same thing: watching how work really happens, noticing where things could be more efficient, and building something that makes the job easier.
These days that means engineering AI infrastructure: a schema-driven tool server that gives language models safe, validated access to live data, agentic workflows that select tools and preserve state, and deterministic pipelines that turn raw field telemetry into something a model can trust. AI-assisted development is my default workflow — I use AI to build AI, and it shows in how fast things ship.
I'm also deepening the foundation formally: currently pursuing my PhD in Artificial Intelligence, because the systems I want to build next deserve more than intuition.
I care about useful things. Tools that answer the question. Systems that make the next step obvious. Technology that feels less like noise and more like assistance.
Instructional systems design came first — launching training sites and learning how people genuinely adopt software.
A beauty business, built from zero to 140+ clients — the brand, the booking flow, the retention system, all designed in-house.
Project leadership across 50+ accounts taught the value of looking closely — one careful audit returned $100K to the business.
Today it's AI infrastructure — a 31-tool server giving models validated access to live data, agentic workflows in production, and AI-assisted development as the daily default.
Currently pursuing a doctorate in Artificial Intelligence. The next systems deserve rigor, not just instinct.
I like the moment when the fog clears. When the messy spreadsheet becomes a clean answer. When a dashboard finally shows the thing everyone kept asking about. When a process that used to take hours suddenly takes minutes.
I'm interested in AI that actually helps — tools that understand the work, check the data, and make decisions easier without adding more noise.
Every process has one. Understanding how things work — then making them quicker, lighter, easier — is the part that feels like play.
Designing the guide, the walkthrough, the aha. Teaching is design at its warmest — it's where everything built comes alive.
A friendly little helper, hand-stitched to answer questions about my work. Go ahead — it likes questions. (The full AI systems live in the projects below.)
Real projects — the code, the demos, the stories. Click into whichever kind of proof you like best.
Every job scored on health and next action — hundreds of moving pieces, one clear and trusted view.
click for the story ↓Leadership wanted one trustworthy answer to "how are the jobs doing?" — available any moment, no meeting required.
React + Next.js over a deterministic scoring engine — operational telemetry ingested, validated, and ranked so risk surfaces before it costs money.
312 jobs · 94% on-time
Status meetings became status glances.
A Python/FastMCP service exposing 31 schema-enforced tools to LLM orchestrators — validated, stateful, fully logged.
click for the story ↓Language models are only as trustworthy as their context. The system needed verifiable, auditable access to live operational data.
A Python microservice on FastMCP: strict JSON schema enforcement on every tool contract, source-freshness validation at runtime, and end-to-end execution logging.
31 contracts · 100% auditable
Every model answer traceable to a current, validated source.
A deterministic ingestion pipeline: GPS-verified time capture validated, structured, and transformed into accounting-ready records.
click for the story ↓Field time for a 100+ contractor fleet deserved a smooth path from jobsite to accounting.
Spatial validation at capture (GPS bounds checking), rule-based payload enrichment (per-diem auto-tagging), and structured export to accounting — clean data in, clean data out.
$198K processed · 92% approved
Weekly invoicing, now a non-event.
A readiness engine — crews mobilize informed, equipped, and cleared.
click for the story ↓Mobilization day goes smoothest when every box is checked before the trucks roll.
A stage-gated readiness funnel with a 17-truck fleet calculator, telemetry, and weather context built in.
7 segments · 1 green light
Crews arrive ready. Day one runs like day ten.
The whole job in a foreman's pocket — status, risk, maps, work orders.
click for the story ↓Foremen work on their feet. The full picture should travel with them.
A fast, read-only React view — job status, material context, maps — that loads beautifully on any phone.
1 pocket · 0 callbacks
The answer is already in hand.
Briefs and reviews that write, design, and send themselves.
click for the story ↓Leadership reads best when the right story arrives at the right rhythm — daily, weekly, monthly.
Apps Script + Gmail HTML, designed like a publication: clear hierarchy, real signals, scheduled delivery.
4 cadences · 0 manual hours
It sends at 6am. People actually read it.
Try help, story, numbers, kreyol — or, if you're hiring, sudo hire-lonna.