Architecting scalable machine learning infrastructure and secure, data-driven platforms.
- 🧠 Current Core: Optimizing LLM architecture, real-time model serving, and KV cache management.
- ⚙️ Architecture: Building secure, high-throughput microservices and memory-safe feature engineering pipelines.
- 🤝 Collaboration: Actively building and refining AI-driven platforms with robust role-based access controls.
- 📬 Contact: mhlongosihle49@gmail.com
- 💼 Professional Profile: LinkedIn | Portfolio
| Project | Description | Stack |
|---|---|---|
| 📊 Behavioral Forecasting Engine | Predictive customer modeling pipeline processing 18+ million transaction records with strict temporal discipline. | Python, XGBoost, PostgreSQL |
| 🛡️ Guardrail Sentinel | Real-time prompt injection detection and security auditing tailored for business AI agents. | Python, FastAPI, Vector DBs |
| 💼 JobIQ | Production-ready AI interview simulator and automated scoring engine built and deployed in a 24-hour sprint. | Azure, Cosmos DB, OpenAI |
- Optimize for memory safety and real-time inference bottlenecks over raw feature accumulation.
- Security and role-based governance are baseline requirements, not afterthoughts.
- Data integrity dictates system design.
- Based in South Africa, building for global scale.