06/26/2026 updated
100 % available
Machine Learning & Deep Learning Engineer | Time-Series & Computer Vision | PhD, TU Delft
Delft, Netherlands
Only remote
PhD, AI for Predictive Maintenance, Delft University of Technology (2021-2025)About me
Machine learning engineer with a PhD from TU Delft who designs, trains, and ships deep-learning models for high-stakes, real-world data. I have built and deployed CNN, LSTM, attention-based, and diffusion models, deep reinforcement learning systems across time-series, image, and tabular data.
LangChainData AnalysisPython (Programming Language)Predictive AnalyticsAzure Machine LearningSensor FusionDeep LearningMachine Learning OperationsFeature Extraction
Co-Founder & CEO / Lead ML Engineer (Prognora)
• Designed, trained, and deployed the deep-learning models behind a production ML platform, owning architecture and evaluation.
• Built the backend of the ML decision-support platform using FastAPI, Pydantic, Docker, and MLflow.
• Built end-to-end training and inference pipelines on multivariate sensor, image, and tabular data.
• Set the technical direction for model development across 6 concurrent AI projects.
• Drove R&D on neural network architectures and uncertainty-aware deep learning.
• Applied LLM-based tooling (Claude Code) in the engineering workflow.
PhD Researcher – Deep Learning (Delft University of Technology)
• Built CNN, LSTM, and attention-based models for diagnostics and prognostics on multivariate time-series, image, and tabular data.
• Developed a fine-tuned conditional diffusion model for generative image synthesis, and interpretable neural networks with constrained weights.
• Designed deep reinforcement learning agents for sequential decisionmaking under uncertainty.
• Implemented Bayesian deep learning for aleatoric and epistemic uncertainty estimation.
• 8 peer-reviewed publications; BestWork Award at 2nd Int. Conf. on CBM in Aerospace.
Services
• Neural Network Development. Architecture design, training, and evaluation of CNNs, sequence, and attention models.
• Generative & Diffusion Models. Conditional diffusion and generative modeling for synthesis and augmentation.
• Deep Reinforcement Learning. Agents and policies for sequential decision-making.
• Uncertainty-Aware Deep Learning. Bayesian methods and calibrated, reliable predictions.
• Research-to-Production. Training and inference pipelines from prototype to deployment.
Data Preprocessing & Cleaning
Expertise across heterogeneous sensor sources including time series, images, and tabular data. Includes statistical analysis, visualization of cleaned signals, and multi-sensor fusion across different modalities such as vibration with strain and acoustic emission with fibre optic.
Feature Extraction
Principled selection and construction of features from cleaned sensor data, producing inputs ready for diagnostic and prognostic ML models. Covers handcrafted, signal-processing-based, and learned representations.
End-to-End AI Deployment
Building end-to-end pipelines from sensor ingestion to decision-support platforms, including model validation and deployment of condition monitoring and predictive maintenance pipelines for industrial pilot engagements.
Multi-Sensor Data Fusion
Fusion of data across different sensor modalities such as acoustic emission, fibre optic strain, and piezoelectric sensors for damage detection in composite aerospace panels.
Azure AI & Cloud Platforms
Microsoft Certified in Azure AI Fundamentals. Experience accelerating end-to-end data science workflows using NVIDIA tools and cloud-based platforms.
Languages
EnglishFluentFrenchGoodGreekNative speaker
Project history
Founded and leads a deep-tech startup commercializing peer-reviewed PHM/SHM research from academic methods into production-grade ML systems for industrial assets. Directs technical architecture, model validation, and deployment of condition monitoring and predictive maintenance pipelines. Leads commercial strategy, customer discovery, and go-to-market across aerospace, wind energy, defence, and adjacent heavy-asset industries. Builds and leads a multidisciplinary founding team spanning ML research, software engineering, and domain expertise.
End-to-end doctoral research covering raw sensor data, fault detection, RUL estimation, and maintenance scheduling. Developed diagnostic and prognostic deep learning models on multivariate time-series, image, and tabular data for composite aerospace structures. Designed Post-Prognosis Decision Making frameworks using Deep Reinforcement Learning under uncertainty and imperfect maintenance assumptions. Affiliated with the Center of Excellence in AI for Structures, Prognostics & Health Management (AISPHM). Received Best Work Award at the 2nd International Conference on Condition-Based Maintenance in Aerospace, awarded by KLM.
Developed neural network models for vibration-based damage detection on aerospace structures under assembly uncertainty. MSc thesis graded 10/10, which served as the foundation for subsequent PhD research direction.