Logistics / B2B SaaS
WithVUS — putting a waste-collection fleet on one dispatch board
WithVUS is a South Korean waste-collection business that came to Aisahub needing an operations platform, and I built it. It covers the full operational surface of a collection company — trucks, drivers, routes, schedules, customers, pickup and delivery locations, and job completion tracking.
- Role
- Software Engineer — full-stack, feature ownership from schema to UI
- Company
- Aisahub, Seoul, South Korea (working remotely from Jakarta)
- Users
- Dispatch admins at Korean waste-collection companies
- Focus
- Calendar dispatching · Mapbox routing & live tracking · Compliance documents
- Timeline
- July 2025 — present
The problem: a fleet is a scheduling problem wearing a map
A waste-collection company looks simple from outside — trucks go out, waste comes back. Inside, it is a dense web of constraints. Every job needs a truck that can carry the right material, a driver licensed to drive it, a route that makes geographic sense, a customer site to collect from, and a disposal site to deliver to. Change any one of those and the rest may stop being valid.
The hard part is that this all has to be planned ahead of the day it happens, and then survive contact with reality: a truck breaks down, a customer reschedules, a site closes early. So the data model has to represent both the plan and what actually occurred, without the two collapsing into each other.
Most of the engineering effort here isn't in any single screen. It's in modelling entities that all reference each other, then keeping every view consistent as they change.
Calendar-based dispatching
The core of the product is a calendar where admins schedule future work. Rather than filling in a form per job, a dispatcher works the way they already think — on a date grid — and assigns the moving parts to each planned job:
- Drivers — who is working, and are they available that day
- Trucks — which vehicle, and is it suitable for the material
- Routes — the path the job follows
- Pickup and delivery locations — where waste is collected and where it goes
A planned job therefore has to exist as a record before anyone drives anywhere, carry its assignments, and later reflect whether the work was completed — which is what job completion tracking depends on.
Why assignment is harder than it looks
Every assignment is a potential conflict. The same truck cannot be on two jobs at once; a driver's day has a shape. In a domain like this, the gap between a tool people trust and a tool they work around tends to be how early those clashes become visible to whoever is doing the scheduling.
Maps: routing and real-time tracking
Dispatch is inherently spatial, so the platform includes interactive maps built on Mapbox with custom routing logic layered on top, plus real-time tracking integration so admins can see where trucks and drivers actually are.
Two distinct problems live here. Routing is a planning question — given these stops, what path should the truck take, and how do we express that in a way the rest of the system can store and re-use. Live tracking is a streaming question — positions arrive continuously and the map has to stay smooth and readable while they do, without turning every update into a re-render of the world.
Getting the map to feel immediate is largely about deciding what not to redraw. The interesting work is in the update path, not the initial render.
The unglamorous half: compliance documents
Waste handling is regulated, and a collection company's obligations don't end when the truck comes back. A meaningful part of my work has been the business-critical document workflows that sit behind the operational layer:
- Weighbridge documents — the recorded weight of what was collected and delivered
- Invoices — billing derived from work that actually happened
- Tax-related records
- Government-related operational documents
These features rarely make it into a product demo, but they are the reason customers can use the software as their system of record rather than as a scheduling toy alongside their real paperwork. They also impose the strictest correctness requirements in the product: an invoice that disagrees with a weighbridge record is not a cosmetic bug.
What I took from it
WithVUS is the kind of system where domain modelling dominates. The frameworks are ordinary; the difficulty is in representing a real operational business honestly enough that the software stays correct when the day goes sideways. Building it for Korean clients — remotely, across a two-hour timezone gap from Jakarta — made the modelling decisions concrete rather than theoretical.
Hiring a full-stack engineer who ships?
Jakarta (GMT+7), fluent in remote-first work across timezones. Usually replies within a day.