AI Strategist & Machine Learning Engineer
PhD in Applied Mathematics · Bayesian & Probabilistic Machine Learning
Founder, TeraSystemsAI · Research in Healthcare & Security
I design and build reproducible, uncertainty-aware machine learning systems for real-world decision-making, particularly in high-stakes domains such as fraud detection, customer retention, and healthcare.
My work bridges peer-reviewed research and production-grade ML pipelines, with a strong emphasis on interpretability, evaluation rigor, and risk-aware modeling.
| Title | Journal | Year | DOI |
|---|---|---|---|
| Bayesian RAG: Uncertainty-Aware Retrieval for Reliable Financial Question Answering | Frontiers in Artificial Intelligence | 2026 | https://doi.org/10.3389/frai.2025.1668172 |
| Hybrid Naïve Bayes Models for Scam Detection | IEEE Access | 2025 | https://doi.org/10.1109/access.2025.3569216 |
| Enhancing Autonomous Systems with Bayesian Neural Networks | Frontiers in Built Environment | 2025 | https://doi.org/10.3389/fbuil.2025.1597255 |
| Application of Bayesian Neural Networks in Healthcare | Machine Learning and Knowledge Extraction | 2024 | https://doi.org/10.3390/make6040127 |
End-to-end machine learning pipeline for customer churn risk prediction using real-world tabular data, with a focus on reproducibility, leakage control, evaluation discipline, and explainability.
https://github.com/lebede-ngartera/customer-churn-risk-ml
Decision-intelligence project integrating probabilistic demand forecasting, constrained optimization, and Monte Carlo risk evaluation for supply chain planning under uncertainty.
The project demonstrates how uncertainty-aware forecasts (P50 vs P90) translate into materially different operational decisions when subject to inventory, capacity, and budget constraints, and how CVaR-based stress testing reveals tail risk invisible to mean-based metrics.
Focus areas:
Forecast-to-decision linkage (forecast quantiles → actions)
Optimization under operational constraints
Risk-aware evaluation using Monte Carlo simulation
Executive-style decision memos communicating cost–service–risk tradeoffs
https://github.com/lebede-ngartera/supply-chain-decision-intelligence
Implementation and evaluation of hybrid Naïve Bayes–based models for real-world scam detection, derived from peer-reviewed research.
- Bayesian generative modeling under data sparsity
- Cost-sensitive evaluation in highly imbalanced settings
- Uncertainty-aware decision thresholds
Derived from: Hybrid Naïve Bayes Models for Scam Detection (IEEE Access, 2025)
Repository: (in progress)
- Research defines what is theoretically sound
- Engineering determines what is deployable
- Evaluation decides what is trustworthy
I focus on making tradeoffs explicit and uncertainty visible, rather than optimizing single-point metrics in isolation.
- Fraud & scam detection
- Customer behavior modeling
- Bayesian & probabilistic machine learning
- Risk-aware AI systems
- Interpretable ML in regulated environments