Applied AI & research software for engineering and environmental systems
Digital twins · anomaly detection · sensor-data QA/QC · scientific Python · environmental and industrial systems
I'm a Research Fellow and Data Scientist working at the interface of machine learning, sensor data, and engineering/environmental systems, based at the University of Edinburgh. My work turns complex engineering and environmental data into reproducible tools for monitoring, validation, anomaly detection and decision support.
I'm especially interested in applied AI for sensor-rich physical systems — structural testing, hydrology and hydraulics, and environmental monitoring — where the hard part is usually the data pipeline and validation, not the model.
The best starting points are:
- Scientific ML / geomorphology:
meander-morphology-classifier - Digital twins / industrial AI:
synthetic-hydraulic-digital-twin-demo - Applied ML / anomaly detection:
audio-anomaly-detection-structural-testing - Environmental monitoring / sensor QA/QC:
urban-drainage-sensor-data-toolkit - Scientific modelling / hydrology:
LDSFL_Meander - Engineering data QA/QC:
tdms-sync-checker
Together, these projects show how I approach applied AI beyond model fitting: data quality, reproducibility, validation boundaries, and honest documentation of assumptions and limitations.
- Urban drainage telemetry QA/QC demo: Launch demo · Repository
- Meander morphology classifier demo: Launch demo · Repository
These demos use public-safe synthetic or example data so the workflows can be explored without installing the repositories locally.
| Project | Area | What it demonstrates |
|---|---|---|
Meander Morphology Classifier |
Scientific ML / geomorphology | CWT spectra, autoencoder latent spaces, clustering, Streamlit GUI, Zenodo-linked models, reproducible meander-bend classification workflows, and a one-click demo |
Hydraulic Digital Twin |
Digital twins / industrial AI | Synthetic digital-twin workflow for hydraulic systems, from generated sensor data to anomaly detection, state classification and decision-support reports |
Structural Audio Anomaly Detection |
Applied ML / anomaly detection | Audio-based anomaly screening for structural test campaigns |
Urban Drainage Sensor Data Toolkit |
Environmental infrastructure / sensor QA/QC | Public-safe telemetry QA/QC, synthetic drainage-monitoring data, automated reports, anomaly screening, synthetic monitoring-map outputs, and a one-click interactive demo |
LDSFL Meander |
Scientific computing / hydrology | Reduced morphodynamic modelling for river meander evolution |
TDMS Sync Checker |
Engineering data QA/QC | Synchronisation and integrity checks for multi-channel TDMS sensor data |
- Full-scale tidal blade testing:
tidal-blade-test-analysis— public-safe research-software workflows for full-scale composite tidal-blade structural-test data, including fatigue summaries, natural-frequency helpers and applied-AI screening.
These repositories complement the main pinned projects by showing how I convert private engineering data into public-safe, reproducible workflows.
-
Remote sensing / environmental monitoring:
strandings_from_space— collaborative research software for VHR satellite-image pre-processing, annotation and observer-count comparison for cetacean strandings. My fork is available atsergioald/strandings_from_space. -
Open-source research software / deep learning:
GeoOcean/BlueMath_tk— upstream contributions to the deep-learning autoencoder module of a climate-data analysis toolkit.
- Applied AI: anomaly detection, classification, time-series and signal features, model validation
- Engineering data: sensor networks, TDMS files, synchronisation diagnostics, data-quality checks
- Environmental data: remote-sensing workflows, hydrology, hydraulic modelling, urban drainage
- Scientific ML: autoencoders, latent spaces, clustering, spectral features, river morphology
- Scientific Python: NumPy, pandas, SciPy, Matplotlib, scikit-learn
- Research software: reproducible workflows, command-line tools, GUI tools, documentation, tests
I try to make repositories useful as engineering and research artefacts, not just as code. Where possible, projects include a clear problem statement, installation and usage instructions, tests, and an explicit account of what is validated and what is not. This matters most when real industrial, environmental or research data cannot be published — in those cases, the repository is built around a synthetic or public-safe version that still demonstrates the actual workflow.
I'm interested in applied AI, research software, digital twins, sensor-data QA/QC, and environmental/industrial monitoring.
- Portfolio: [sergioald.github.io](https drainage
- Scientific ML: autoencoders, latent spaces, clustering, spectral features, river morphology
- Scientific Python: NumPy, pandas, SciPy, Matplotlib, scikit-learn
- Research software: reproducible workflows, command-line tools, GUI tools, documentation, tests
I try to make repositories useful as engineering and research artefacts, not just as code. Where possible, projects include a clear problem statement, installation and usage instructions, tests, and://sergioald.github.io/)
- GitHub: @sergioald
- LinkedIn: Sergio Lopez Dubon
- Academic profile: University of Edinburgh Research Explorer
- Publications: Google Scholar
- ORCID: 0000-0003-0663-607X