Dingaan Mahlatse
Machethe
Data Scientist · Data Engineer · AI/ML Engineer · AI Automation Developer · Cloud Security Researcher
Building production AI systems that turn raw institutional data into decisions someone can act on — multi-agent pipelines with a human approval gate, ML models, and dashboards over a live school of 950+ learners.
About
Builder. Engineer. Educator. Researcher.
Data Scientist, AI/ML Engineer, and Azure Data Engineer with 9+ years of applied experience building production data systems in the education sector. I built a full AI-powered EdTech ecosystem from scratch — serving 950+ learners — while serving as Head of Department (Science, Mathematics & Technology) at a South African public high school. Dual MSc candidate in Data Science (University of East London) and Cybersecurity — Cloud Security Architecture (EC-Council University).
I designed and built the entire EduAnalytics + EduPortal stack: a Python/FastAPI backend with five AI agents (ReAct tool-calling, RAG, multi-provider LLM), a production RandomForest classifier (85%+ accuracy, 20 engineered features), XGBoost end-of-year predictor, LSTM term forecaster, and a Next.js 16 portal with real-time AI chatbot, RBAC, and Supabase PostgreSQL — integrated via a 10-minute auto-sync bridge.
MSc Cybersecurity research (EC-Council University, ECCU500) produced a peer-reviewed paper on ML-based intrusion detection for cloud network security within a Zero Trust Architecture framework. Five ML models were benchmarked on NSL-KDD (148k records): LSTM achieved 98.1% accuracy and 1.8% FPR — the top result across all models. The framework maps each model to one of five NIST SP 800-207 ZTA pillars across a six-phase implementation roadmap.
As head of SKYLearn-Innovation NPO, I mentor community learners — high-school students and a university tutor — through hands-on data science projects. The flagship outcome is ZA-STEM Insights, a 10-page live analytics platform on 10 years of South African matric data: multi-model ML (XGBoost, SHAP explainability), K-Means clustering, Holt ETS forecasting, hypothesis testing, and a live SQL layer — deployed publicly on Streamlit Cloud.
Proficient in the full ML lifecycle: data ingestion, feature engineering, model development, FastAPI deployment, and MLOps monitoring. Actively building RAG pipelines, LangChain/LangGraph agent frameworks, and vector search capabilities. Available immediately for Data Science, Data Engineering, AI/ML, AI Automation, or Cloud Security Engineering roles — nationwide, remote or hybrid.
Applied Experience
Data Scientist & AI Systems Engineer
Mahlontebe Secondary School · Self-initiated applied research role
- Built EduAnalytics: 5-phase AI agent pipeline (ReAct, RAG, multi-LLM) + RandomForest 85%+ classifier + LSTM term forecaster — FastAPI + Docker + MLflow
- Built EduPortal: Next.js 16 portal carrying a five-agent accountability layer — management-plan follow-up, audit accountability, due-date escalation, leadership briefing, sampling without replacement — behind a human approval gate
- Built the school’s WhatsApp integration both ways: Meta Graph API out, HMAC-SHA256 signed webhook in, with consent re-checked at send time under POPIA
- Designed full SASAMS ETL pipeline: CSV ingestion → PostgreSQL → ML feature engineering — end-to-end data ownership
- Built Azure Medallion pipeline: ADF → ADLS Gen2 → Databricks PySpark → Synapse Analytics (100M+ records, star schema)
ML Research Engineer — Cybersecurity
EC-Council University · MSc Research (ECCU500)
- Benchmarked 5 ML models (RF, SVM, LSTM, Autoencoder, XGBoost) on NSL-KDD (148k records) — LSTM: 98.1% accuracy, 1.8% FPR
- Designed ZTA framework mapping each model to NIST SP 800-207 pillars across a 6-phase implementation roadmap
Founder & Lead Data Scientist
SKYLearn-Innovation NPO · Registered South African NPO
- Led team of 5 (4 high-school learners + 1 university tutor) to build ZA-STEM Insights: 9-page Streamlit analytics platform on 10 years of SA matric data
- Delivered XGBoost + SHAP explainability, K-Means clustering, Holt ETS forecasting, hypothesis testing — live on Streamlit Cloud
Publications
EC-Council University Cyber Journal · ECCU 520 Research Project
Peer-reviewed research on the two injection classes that still dominate the OWASP Top Ten, examined in the cloud-hosted e-commerce setting where they do the most commercial damage.
Education
MSc Data Science
University of East London (UEL)
u3276087@uel.ac.uk
MSc Cybersecurity — Cloud Security Architecture
EC-Council University (ECCU)
dingaan.mahlatse@eccu.edu
Term 3 — In Progress
Postgraduate Diploma in Data Science
Regenesys Business School
Completed
Prior Education
Postgraduate Diploma in Public Management
Regenesys Business School
PGCE — FET & Senior Phase (Mathematics & Physical Sciences)
University of Limpopo
BSc — Physics, Chemistry & Mathematics
University of Limpopo
National Senior Certificate (Umalusi)
Grade 12
Projects
Production systems, not side projects.
Every project below is deployed, serving real users, or backed by working code — not slides.
Live Analytics Demo
What the system actually produces.
The EduAnalytics multi-agent pipeline generates these insights automatically each term — urgency scoring, at-risk identification, trend detection, and stakeholder communications.
At-Risk Learners by Grade
RandomForest classifier · 0.5 probability threshold
Term 1 vs Term 2 Pass Rates
TrendAgent · pass-rate delta detection per subject
AI Agent Urgency Scores — Subject Priority Ranking
PrioritiserAgent · composite score: pass rate, at-risk count, trend severity, teacher workload
🔴 ≥ 70 = Critical intervention · 🟠 50–69 = Monitor closely · 🟢 < 50 = Stable
Skills
Production-proven, not tutorial-level.
✦ = sandbox / structured self-study exposure. All other items reflect applied production experience.
Data Engineering
AI / ML & MLOps
Full-Stack & BI
Cloud & Security
Contact
Let's build something together.
Remote or hybrid. Open to contract, fixed-term, or permanent roles. I bring production-ready systems, not just portfolio projects.
Email (Primary)
machethedm@gmail.com
Phone
+27 83 875 1445
linkedin.com/in/dingaan-mahlatse-machethe
GitHub
github.com/machetheDM
Live Portal
mahlontebe.org.za
Academic Contacts
Academic emails are for institutional correspondence. Primary contact is Gmail.
This portfolio is hosted on the same Next.js 16 + Vercel infrastructure serving EduPortal, a production platform with 950+ users.