Michele E. J. Maestrini
Data Scientist | Predictive Analytics | Machine Learning | Time-Series Analysis
Professional Summary
Data Scientist focused on turning complex operational and engineering data into actionable decisions through machine learning, predictive analytics, and statistical modelling. Michele combines engineering training, large-scale operational experience, and modern data science techniques to improve reliability, optimise resources, and support decision-making in complex environments.
Experience spans predictive maintenance, time-series modelling, machine learning, operational analytics, and deployment-oriented workflows. Current project evidence includes FusionCore, a NASA C-MAPSS prognostics programme for Remaining Useful Life estimation, and an MSc research thesis in applied few-shot learning.
Core Competences
The current stack spans Python, SQL, Excel, pandas, NumPy, scikit-learn, XGBoost, TensorFlow/Keras, anomaly detection, model calibration, zero-leakage validation, RUL estimation, Weibull analysis, fleet-health analytics, and PSI drift detection.
Deployment and delivery tooling includes MLflow, DVC, FastAPI, Pydantic, Docker, Kubernetes, GitHub Actions CI/CD, AWS ECR/EKS, Gradio, Jupyter, Git/GitHub, and VS Code.
The detailed breakdown of tools, methods, and competencies now sits on its own page so the main Curriculum Vitae view stays focused on profile, evidence, and career direction.
Education
MSc Data Science
Birkbeck, University of London
Merit award, with a few-shot learning dissertation and a 93% distinction-level result in Neural Networks and Deep Learning.
BEng Civil Engineering
University of Westminster
Foundation in structural analysis, impact loading analysis, and engineering mathematics, providing the physical intuition behind the applied AI work.
Selected Project Focus
Predictive Maintenance & RUL Modelling
NASA C-MAPSS PHM project focused on RUL estimation, reliability analytics, and maintenance decision support.
- Built a zero-leakage turbofan RUL pipeline using 91 physics-aware features across thermodynamic, kinematic, fatigue, compressor-efficiency, and risk indicators.
- Evaluated XGBoost across 707 held-out engines, achieving RMSE 14.85, NASA Score 4,336, and Critical-band F2 0.9339.
- Added conformal intervals and calibrated risk bands, then benchmarked PiNet hybrid deep learning against XGBoost for operational model selection.
Applied Few-Shot Learning
Built a MobileNetV2/Siamese few-shot learning model using triplet loss and Bayesian optimisation, achieving 98.75% test accuracy and 98.77% recall under data-constrained conditions.
Professional Experience Highlights
Commercial Operations & Logistics
Reduced operational costs by 10% across 12 teams through cost, staffing, and workflow analysis. Coordinated Christie’s Paris-to-Monaco auction logistics for approximately £70m in consignment value, involving three teams, three articulated trucks, and police-escorted delivery.
Site Engineer | Design Management Group
Delivered five structural work packages across steel-frame, reinforced concrete, masonry, and lift-shaft strengthening works. Coordinated contractors, fabricators, building control, and approximately ten site workers while producing calculations, drawings, method statements, and sequencing guidance.
Additional Strengths
Tooling & Workflow
- MLflow and DVC experiment and artefact workflows
- FastAPI and Pydantic serving interfaces
- Docker, Kubernetes, and GitHub Actions CI/CD
- AWS deployment using ECR and EKS
Analytical & Interpersonal
- Large-scale operational analysis
- Scheduling, routing, and resource allocation
- Reliability and maintenance decision support
- Engineering-grounded problem solving
- Stakeholder and contractor coordination
