Operations · Networks · Practical AI
A little curiosity.
A more useful answer.
I’m Felipe. My work connects operations and network optimization. Here, I explore how practical AI and small, transparent tools can make complex decisions easier to understand.
Try a small experimentA personal home, with a small lab attached.
The little lab / Experiment 01
Synthetic data · EducationalWhen is a second hub
worth it?
Shorter journeys. Another fixed cost.
Move the sliders and see where the balance changes in a fictional delivery network.
Set the scene
Every change stays in your browser.
No uploads. No live AI calls.
Two hubs cost $383 more per day.
They bring 45 percentage points more parcels within the 2-day target.
All data is invented. Outputs are educational, with no employer data or commercial validation.
Download synthetic dataLook under the hood Assumptions, calculations & limits
A deliberately small model
Four fictional regions keep the same demand shares. One hub serves all regions from West. With two hubs, Ash and Brook stay with West; Cove and Dune move to East. Both hubs stay open even at zero demand.
Delivery time = 1 processing day + distance ÷ 400 km per day. A parcel meets the target when this time is less than or equal to the chosen deadline. This is a simplified deterministic rule, not a delivery forecast.
Daily cost = fixed cost + (parcels × handling cost) + (parcels × weighted distance × $0.007 × transport multiplier). Transport uses one-way distance; return trips are excluded.
Only the network sketch changes when you select a hub layout. Both cost and service comparisons always remain visible.
The numbers behind the sketch
| Region | Share | West | East |
|---|---|---|---|
| Ash | 30% | 120 km | 760 km |
| Brook | 25% | 300 km | 520 km |
| Cove | 25% | 620 km | 180 km |
| Dune | 20% | 900 km | 140 km |
One hub: $900 fixed/day + $1.80/parcel handling. Two hubs: $2,100 fixed/day + $2.00/parcel handling. Weighted distance: 446 km vs. 184 km.
At this transport rate, costs cross at about 734 parcels/day.
Left out on purpose: capacity, inventory, congestion, variability, emissions, taxes, hub setup costs, and route optimization. Fractional parcel shares are allowed. Dollar amounts are rounded only for display. At zero demand, service percentage is undefined.
Selected work
A small collection, on purposeSmall things you can try.
01The second-hub question
Interactive prototypeA transparent cost-and-service experiment. Built for exploration, using only fictional inputs.
Toothpaste Plant
Historical inputs · Synthetic simulationExplore six production lines, capacity and scheduling using dated historical figures and explicitly synthetic operating scenarios.
Small experiments, with their inputs, assumptions and limits open to inspection.
A note on this space
Make the thinking visible.
I’m interested in the space between a complicated system and a practical decision. Operations and network optimization are part of that lens; experimenting with AI is another.
This is a place to turn a question into something you can inspect, adjust, and learn from. Small experiments. Clear assumptions. Room to be wrong.
A short introduction
A short hello, made with my AI avatar. 50 seconds · Transcript below.
Read the transcript
Hi, I’m Felipe Vilhena. Welcome. I’ve spent more than twenty-five years in supply chain, including procurement leadership at Church & Dwight and experience at Colgate-Palmolive.
I’m interested in how better supply chain decisions can make a business stronger. What is driving a problem? Which trade-offs really matter? And what changes when you try a different approach?
That’s the thinking behind this website. Here, I’m bringing my supply chain experience together with my curiosity about artificial intelligence. You’ll find interactive demos, practical tools, and experiments you can explore for yourself.
The supply chain demos use synthetic data. These are personal projects, built to explore ideas without sharing company information. Take a look around, try a scenario, and see what changes. Thanks for being here.