Mirakl is a French unicorn and global leader in enterprise marketplace solutions. Its technology enables leading retailers, manufacturers, and B2B companies to launch and scale online marketplaces efficiently. Mirakl powers hundreds of marketplaces worldwide, helping organizations expand product offerings, improve logistics, and create seamless e-commerce experiences.
Joining Mirakl
I joined Mirakl as a Data Scientist in December 2024, where I had the opportunity to work on one of the company’s key machine learning initiatives: estimating delivery dates for marketplace orders.
This project allowed me to apply everything I had learned so far, from model design to full-scale production deployment.
Project Overview
The goal was to build a delivery date estimation model capable of predicting when each order would reach the customer.
The project covered the entire machine learning lifecycle, including:
- Data collection and preprocessing
- Time series model design and proof of concept
- Validation of POC with the business team
- Pipeline orchestration, production deployment, testing
- Performance monitoring
From Beta to General Release
The beta program launched in July 2025 with three pilot clients, providing estimated delivery dates for over 50,000 orders per week.
After several months of testing and monitoring, we released the feature to general availability in October 2025.
What I Learned
Working on this project was both technically challenging and deeply rewarding.
Here are some of my key takeaways:
- Cross-functional collaboration — I worked closely with SREs, data engineers, product managers, developers, and BI analysts.
- Modern data infrastructure — Everything was managed as code, leveraging tools like Spark, Databricks, Airflow, and MLflow.
- End-to-end ownership — From the first notebook to production pipelines and monitoring dashboards.
Final Thoughts
This project was an amazing experience in building a production-grade machine learning system from scratch. I learned what it truly means to bring a model into production: handling reliability, scalability, and real-world constraints. I am deeply grateful to have had the opportunity to work on this project.
That said, completing this project also helped me realize something deeper about myself: while I enjoyed the engineering and operational aspects, my real passion lies in research: exploring new methods, pushing boundaries, and solving major scientific problems. This realization led me to transition back toward research-oriented projects after the algorithm was released.
