This platform is itself one of those systems: a database-backed, full-stack application, not a static portfolio describing one. Every rule it enforces — a publish gate that checks client approval server-side, an audit log every admin action writes to, an AI agent with no privileged write path — is the same discipline applied to client and personal work alike.
About
Kurhula Success Maluleke
Software & AI Engineer
- Final-year BSc Computer Science & Mathematics student · North-West University
I’m a final-year Computer Science and Mathematics student building production-grade systems — full-stack web, AI integration, and enterprise automation — across fintech, EdTech, GovTech, and SaaS. Architecture and design come before any code is written; one system gets built at a time, no forward dependencies, no shortcuts taken to hit a date instead of a standard.

- 2 years building
- 11 published systems
- 46 skills proven in code
- 2 companies
In my own words
How I build
The method.
“I build systems disciplined enough to be trusted with real money, real institutions, and real people’s outcomes — engineered in South Africa, held to a global standard.”
- 01
Make It Exist First
Make it exist now; make it beautiful later.
A working version in front of real people beats a perfect one that never ships. This platform went live first, then was redesigned page by page from what using it showed — polish is earned by existing, never a reason to wait.
- 02
Extension Over Modification (EXT-1)
Anything expected to grow lives in data, never in a hard-coded list.
Anything expected to grow — a new project category, a new type of visitor, a new skill — lives in a lookup table or config, never a hard-coded list. A new chapter of the work shouldn't require rebuilding the platform to fit it.
- 03
Smart Not Hard
Buy the commodity, build the differentiated.
Buy the commodity, build the differentiated. Off-the-shelf tools handle what's already a solved problem; real engineering time goes into the parts that actually need building.
- 04
Controlled Imperfection Engineering
Failures made predictable and traceable, not chased into an impossible zero.
Failures are made predictable and traceable, not chased into an impossible zero. Every admin action on this platform writes to an audit log — what goes wrong feeds directly into what gets fixed next, the same discipline an incident produces a runbook.
- 05
Permission Boundaries
No AI acts autonomously on anything that matters.
No AI acts autonomously on anything that matters. The one write path an automated agent can ever trigger here is the same inquiry form a human uses — never a more privileged shortcut.
What I work with
Skills, with proof.
A number shows how many published systems use a skill — open it to see them. The rest are part of his toolkit, not yet in a published system.
Infrastructure4/8 proven
- GitHub
- Vercel
- Sentry
- GitHub Actions
- Better Stack
- Docker
- Prometheus
- Railway
Testing5/6 proven
AI / ML10/11 proven
Languages3/7 proven
- Python
- TypeScript
- SQL
- C++
- Java
- JavaScript
- MATLAB
Database3/6 proven
- PostgreSQL
- Redis
- Neon
- ChromaDB
- MongoDB
- Oracle
Let's build something real.
AIKurhula’s guide
Curious about something? Ask me about him.
