AymenBenkouiten
AI engineer working on anomaly detection, generative models and time series — applied to industrial, medical and telecom data. End-to-end pipelines, from raw signal to a model that holds up in production.
// 02 — WORK
Selected work
Four projects, four different signals. Same playbook: understand the domain, build the pipeline, ship the model.
- P01
CAN Anomaly Detection
Transformer-VAE on raw automotive bus data
CASE - P02
SimBox Fraud Detection
Imbalanced classification + metaheuristic search
CASE - P03
Geo Customer Dashboard
Interactive geographic analytics for decision support
CASE - P04
Parkinson Detection
ML on physiological time series for early diagnosis
CASE
// 03 — APPROACH
How I work
I build end-to-end. Domain framing, pipeline architecture, modelling, evaluation against the metric the business actually cares about. I prefer hard, real-world data over synthetic benchmarks.
I'm drawn to systems where being wrong costs something — safety, fraud, health. That tends to keep the engineering honest.
01
Machine Learning
- Anomaly detection
- Generative models — VAE · Transformer
- Supervised / unsupervised learning
- Time-series modelling
02
Data Science
- Pipeline design — raw to deploy
- Feature engineering on noisy signals
- Imbalanced & adversarial data
- Visualisation & interpretation
03
Stack
- Python · pandas · numpy
- PyTorch · scikit-learn
- Signal processing
- Reproducible data pipelines
// 04 — CONTACT
Got a signal you can't decode?
CDI, freelance, research collaborations. I read every brief. The harder the data, the more interesting the conversation.