Projects
KOS — kitchen operating system
Product and build (solo prototype) · Ongoing
One truth for the whole menu — ingredients, recipes, and costs connected, trusted everywhere.
The real production run — bones to demi-glace. A working prototype and portfolio piece — not a product. It connects ingredients, recipes, and costs into a single dataset a kitchen can trust end to end.
A shop that only sells a feasible pickup
Product and build · 2025
Pre-order for a real Swiss kitchen that only accepted pickups it could fulfil — 10 sandwiches per 15-minute slot, enforced in the path.
Most food sites sell a menu. This one sold a feasible pickup.
I built the pre-order shop for Restaurant National in Wattwil: three sandwiches, drinks, a Thursday/Friday 15-minute lunch slot, pay, collect. No delivery. The kitchen only saw paid orders, grouped by slot.
The constraint was the product. Ten sandwiches per 15-minute slot. Daily SKU caps (chicken 50, pork 30, veggie 30). Same-day cutoff at 10.30am Switzerland. The site checked those limits in the interface, again on the API, and reserved them atomically so two guests could not take the last Fried Chicken. After 10.30am, same-day ordering disappeared.
The kitchen side was the other half: a slot-grouped order view, short spoken IDs, confirmation mail that does not double-send if the payment webhook retries, and a payment flag that could run disabled, demo, or live so we could rehearse without charging guests.
It was also the first stage of what later became KOS. Combining orders and recipes, I could know what needed to be prepped per component — not a list of sandwiches, a list of parts.
This is the only full-cycle recording I have — taken before a polish pass on the shop and the order viewer. The viewer rows are dummy. The site went live for two months; the restaurant was sold and the new owner did not keep the channel.