( Ongoing 🚧 )

AXA - Direct Assurance / Université Paris Cité

Apprenticeship - Machine Learning Engineer

---

Welcome to this repository that captures my year as a Machine Learning Engineer Apprentice at AXA - Direct Assurance.

Overview 🌟

• Company: AXA - Direct Assurance
• Duration: October 2023 - October 2024 (1 year)
• Location: Paris, France
• Role: Machine Learning Engineer / Data Scientist

Objective 🚀

My primary goal is to enhance legacy ML systems by implementing MLOps best practices. This includes the development of automated ML pipelines that incorporate training, feature selection, data drift detection, and model deployment, all based on existing production ML codes.

Core Projects and Responsibilities 🏆

Here's a snapshot of what I've been working on:

1. Development of a Kedro-Based Tool for Monitoring Model Drift on Databricks: Implementing a pipeline in Kedro on Databricks to track model/data drift in Machine Learning models in production.

2. RAG with Longchain for Contextual Querying: Creating a Retrieval Augmented Generation (RAG) model using Longchain to provide contextual responses to email queries within the LLM framework (e.g., GPT, LLAMA).

3. Evaluation of Data/ML Hackathons and Creating Jupyter Notebook Guides: Reviewing Data/ML hackathons and developing Jupyter Notebook guides that provide theoretical insights and practical walkthroughs.

4. Kedro Model Deployment on Databricks: Collaborating with internal ML/Data teams and external service providers to deploy machine learning model training pipelines using Kedro on Databricks.

Tools and Technologies 🛠

During the Apprenticeship, I have been engaged with the following tools and technologies:

• Azure: Used as the primary cloud provider.
• Longchain: Used for enhancing LLMs with the ability to handle contextual inquiries effectively.
• Azure DevOps: Used for code version control and to implement CI/CD pipelines, facilitating automated testing and deployment.
• Kedro: For constructing robust, maintainable, and scalable Data Engineering/ML pipelines.
• Scikit-Learn / Jupyter Notebook / Python: Utilized for developing, testing, and prototyping ML models.
• MLflow: To manage the ML lifecycle, including experimentation, reproducibility, and deployment.
• Azure Databricks: Platform for building and deploying data and ML workflows.

Connect 🌐