Sébastien Chazallet

Okeiro is a health tech startup that publishes software for monitoring transplant patients. The platform assists hospitals and healthcare teams in long-term post-transplant follow-up: centralizing patient data, early detection of complications, and preventing rejection. The domain involves sensitive health data and a direct stake in the quality and continuity of care. The mission took place during a critical phase: securing the platform's production launch.

Data engineering environment on Google Cloud Platform (GCP): Python processing and pipelines, Airflow orchestration, DBT transformations, SQLAlchemy data access, uv for tooling and packaging. Industrialization via Docker and CI/CD. High stakes for the reliability and integrity of the data feeding the platform.

Reinforcement of the data team to meet a critical production launch deadline:

- Design of data migration and processing procedures (recovery, transformation, integrity checks).

- Securing and industrializing ETL pipelines: Airflow orchestration, DBT transformations, robustness, and reusability of processes.

- Quality & Industrialization: implementation/strengthening of tests, CI/CD, and Docker containerization.

- Close collaboration with data and product teams to ensure data integrity at go-live.

Result: More reliable pipelines and mastered processing procedures, supporting a secure production launch.

The BackMarket platform is an online marketplace for refurbished goods.

The team I worked with is based in Barcelona and is responsible for managing an application that allows resellers to list their products for sale and their prices for each country.

To help resellers choose the right price, the application provides information to enable them to position themselves against their competitors, for each product and country. This information can help them adjust their prices to become the active reseller.

This information is mainly compiled by scheduled scripts or SQL queries whose results are placed in specific directories by the data team. Other information falls within the scope of other microservices.

I worked mainly on two microservices, one for managing seller data and the other intended to be consumed by the front-end.

These microservices are built using hexagonal architecture.

The mission consisted of creating specific data models (using SQLAlchemy) and developing the various components needed to integrate data into the microservice's database. This data was accessible by downloading it from GCP and parsing it, or by querying other microservices' APIs.

The development also included performance optimizations on HTTP sessions and the creation of a cache system using Redis.

I also created APIs intended to be consumed by other microservices or by the front-end to expose data to the end-user and participated in building all the internal mechanisms and created components for using data buses to transmit information when changes were made by sellers.

Creation of an application to manage HCERES evaluations, from scratch.

Research organizations are subject to an HCERES evaluation every 5 years. To this end, 8 laboratories have joined forces to create an application to facilitate the collection of this data.

The application must allow, in several phases, the input of all data on laboratory employees, particularly their academic and professional background within the laboratory, but also their research, their responsibilities, and their achievements (recognition indices).

The team working on the project consists of developers from different laboratories, with varying skill levels and heterogeneous testing and production procedures, some using Docker or equivalents, and some not.

I intervened as a technical expert to finalize the application's architecture and deployment procedures, as well as a lead developer to teach and implement development techniques with the team and make choices about third-party modules to use, including Jazzmin.

We were able to develop all the essential aspects of the application in a very short time, imposed by the HCERES evaluation dates, and the application was put into production and used to meet its needs.