Data science undergraduate (CGPA 3.84/4.0) building end-to-end ML systems — XGBoost pipelines with global & local SHAP interpretability, dialect-aware LLM business intelligence, and deployed production apps. Data-Science-Student (CGPA 3.84/4.0) im Bereich Machine Learning — XGBoost-Pipelines mit SHAP-Modell-Erklärbarkeit, LLM-gestützte Business Intelligence und produktive Webanwendungen.
I am a Data Science undergraduate at Albukhary International University (AIU), Malaysia, on a 100% Fully Funded Scholarship. I specialize in building end-to-end machine learning pipelines with rigorous model explainability (SHAP) and deploying production-grade Generative AI applications.
Ich bin ein Data-Science-Student an der Albukhary International University (AIU) in Malaysia mit einem 100 % voll finanzierten Stipendium. Mein Schwerpunkt liegt auf End-to-End-Machine-Learning-Pipelines mit Modell-Erklärbarkeit (SHAP) sowie produktiven Generative-AI-Systemen.
With three deployed projects—including an LLM-powered SQL BI command center (DataPilot AI), an explainable telecom churn dashboard, and a live campus marketplace (AIU Market)—I transform complex datasets into interpretable, high-impact intelligence.
Mit drei produktiven Projekten – darunter ein LLM-gestütztes BI-Kommandozentrum (DataPilot AI), ein erklärbares Churn-Dashboard und ein aktiver Campus-Marktplatz (AIU Market) – verbinde ich technische ML-Algorithmen mit transparenter Entscheidungsfindung.
Built an LLM-powered business intelligence platform converting natural language into dialect-aware SQL across PostgreSQL, MySQL, and SQLite. Engineered AST security guardrails blocking 34/34 SQL-injection attack vectors with sub-50ms query cache latency and automated KPI discovery across 4 business domains. LLM-gestütztes BI-Tool zur Übersetzung natürlicher Sprache in dialektgerechtes SQL. Enthält AST-Sicherheits-Guardrails zur Blockierung von 34/34 SQLi-Angriffsvektoren, sub-50ms-Query-Caching und automatisierte KPI-Plotly-Dashboards über 4 Unternehmensdomänen.
End-to-end churn prediction pipeline on 7,043 telecom customer records with stratified oversampling for class balance. Benchmarked Logistic Regression, Random Forest, and XGBoost (ROC-AUC 0.88, Recall 0.78). Integrated global/local SHAP explainability (beeswarm & waterfall plots) inside a 5-page interactive Streamlit dashboard. End-to-End-Abwanderungsprognose auf 7.043 Telekom-Datensätzen mit stratifiziertem Oversampling. XGBoost erzielte beste Ergebnisse (ROC-AUC 0.88, Recall 0.78). Integrierte globale und lokale SHAP-Erklärbarkeit in einem interaktiven 5-seitigen Streamlit-Dashboard.
Co-built a centralized campus marketplace exclusively for AIU students within a 3-person team, enabling trusted peer-to-peer buying and selling. Owned frontend development in Next.js backed by Supabase/PostgreSQL, reaching 70+ registered users and 25+ active listings within days of launch. Mitentwicklung eines zentralen Campus-Marktplatzes exklusiv für AIU-Studierende in einem 3-Personen-Team. Leitung der Frontend-Entwicklung in Next.js mit Supabase/PostgreSQL-Backend, über 70 registrierte Nutzer und 25+ aktive Angebote in den ersten Tagen.