Hugo Alberola

Development of an intelligent solution to predict the cost and turnover of projects.

Use of Sklearn, LightGBM and XGBoost for training machine learning models for time series prediction.

Retrieval and processing of data in order to create a history that can be used by our models with model versioning via MLflow.

Optimization of hyperparameters with Optuna.

Creation of the API for model inference via Docker and FastAPI

Deployment of the solution on Azure App Service with a CI/CD.

History of predictions on Mongodb via Azure cosmos

The potential delay in the delivery of a package is important and can strongly impact DPD's customers. Requests for information on the delivery status generate costs related to the hosting of support applications. The objective of this project is to anticipate these requests, thanks to all the events that the packages generate (1 billion monthly).

Integration of Spark and Jupyter lab on Kubernetes, querying on Cassandra, development of an anomaly detection model with DBSCAN. Creation of a software platform with Kafka, Cassandra and Spark. model deployment on a streaming flow with Scala and Spark streaming.

Helping web accessibility involves developing accessibility tests. The goal of this project is to automate these tests with Deep Learning techniques.

Scraping of data (Image and textual) on a set of websites. Training and use of computer vision models to subtitle or detect objects in an image (YOLOv3, InceptionV3, Resnet, SSD, Transformer and BERT). Creation of API with FastAPI conteuriser with docker. Deployment of a set of micro service Fast API, Streamlit and Selenium on Azure Kubernetes Service.