We're small on purpose,
and engineer-led by design.
Zephyr wasn't assembled by a holding company. It was built by people who ran switches, chased routes and reconciled rate decks at midnight — and it still is. That shapes how we hire, how we sell, and what happens when you call us.
Five things that decide how we work
Not poster values. These are the ones we've argued about and settled.
Say the difficult thing early
If a route is bad, if a timeline is optimistic, if the thing you're asking for won't do what you hope — we'd rather have that conversation on day one than at go-live. Comfortable answers cost more than honest ones.
Evidence beats confidence
We don't declare a fix working because the logs look right. We wait for real traffic and real records. One outage taught us that lesson properly, and it's now how everything ships.
The person who answers can fix it
No tier-one script, no ticket that gets triaged three times before reaching someone competent. If you call about a route, you get someone who can open the switch and look.
Build it properly the first time
Carrier-grade from the start, even when the shortcut would ship faster. Multi-currency, effective-dated rates, full audit per record — because retrofitting those into a live billing platform is a nightmare nobody deserves.
Small, and staying that way
We're not trying to be the biggest operator in the market. Small means the person who wrote the code is reachable, decisions take hours instead of quarters, and we can say yes to things larger competitors have to decline.
South African, globally useful
Built here, licensed here, data stays here — while terminating A-Z worldwide. Local accountability isn't a limitation on reach; it's the reason enterprise buyers can sign us off.
Frustration is a decent founding principle
Zephyr started because the same problems kept appearing and nobody was fixing them. Carriers renting switches they couldn't extend. Wholesalers losing margin in currency gaps nobody could see. Resellers sold "white-label" that still carried someone else's logo. Call centres paying per seat for software that couldn't hear a local accent.
Every design decision in this platform traces back to something that annoyed us enough to rebuild it.
— How the engineering team describes the roadmapThat's still how features get prioritised. Not by market sizing, but by whether the absence of the thing is costing somebody real money or real sleep.
What it's actually like
You own what you build
No throwing work over a wall. The person who writes a module supports it, gets the calls about it, and fixes it when it misbehaves at 3am.
Nobody is far away
Flat structure, short conversations. If something needs deciding you talk to the person who decides it, usually the same day.
Real systems, real traffic
You'll work on infrastructure carrying 10 million minutes a month. That's a rare thing to get hands on early in a career.
Mistakes get discussed
We've broken production. We write down what happened and why, and the lesson becomes a rule. Nobody gets blamed for a genuine error, only for hiding one.
We hire rarely and carefully
There's no permanent careers pipeline here — when we need someone, it matters enormously who it is. If you've worked in telecoms and recognised yourself in any of the above, write to us anyway. We keep good letters.
- Engineering — SIP, Lua, Python, PostgreSQL, or willingness to learn them properly
- NOC and support — people who can read a trace and stay calm at 2am
- Commercial — wholesale voice experience, or the appetite to acquire it
Want to work with us, or for us?
Either way, the same inbox. We read everything that arrives and reply to the ones that make sense.