Senior Software Engineer
Data systems and
critical applications,
from data to interface
About
I'm a Lyon-based freelance senior software engineer, working for 8+ years on critical systems: the national power grid, medical applications handling data for millions of patients, production AI platforms.
I work the whole stack: large-scale data pipelines (Java, Spring Boot, Apache Spark, Airflow, Kafka), backend services, and web applications with high functional complexity (TypeScript, Angular, React). In practice: between the data schema and the screen that displays it, there is nobody to convince. I carry both sides and the decisions that connect them — which is usually where projects break.
I've used GitHub Copilot since November 2021, in its beta, more than a year before ChatGPT was released. I practice agentic development every day, with Claude Code, Codex and Copilot.
For example at Enedis, an agent proposed reading Kafka messages by parsing every JSON field by hand: fast, but unreadable. I introduced a JSON Schema and a MapStruct mapper: the code became readable, with about four times fewer lines. Even when they technically could, my agents have no access to production or pre-production (Kubernetes cluster, databases): I handle those myself.
I take on long-term missions, autonomously or embedded in a team, on architecture, scaling, technical debt recovery, or team upskilling. If you have a system that needs to hold up, let's talk.
They trusted me
They trusted me
Collaborations as a permanent employee, as a contractor through an IT consultancy, and now as a freelancer
Areas of work
From data architecture to production applications
Areas of work
From data architecture to production applications
High-complexity applications
Business first, code second
Stack & Ecosystem
Streaming & Real-time
Software Craftsmanship
Observability & Metrics
Experience
Designing and scaling data pipelines and critical applications for France's national power grid. I build backend services with Spring Boot and process large data volumes with Apache Spark (Java) and Airflow. I had AI analyze, over one week, the YARN reports of about thirty jobs across 3 environments, to size each job's resource allocation as tightly as possible: execution time down 61% on the DAGs that generate the national Parquet datasets, memory down 87% on our Hadoop cluster. I also use AI (GitHub Copilot, custom agents) to speed up delivery.
Evolved and structured the front-end architecture on a Next.js, React, Tailwind CSS and shadcn/ui stack, following Clean Architecture principles, with business logic isolated in Zustand. Built features, optimized the codebase and supported the team technically.
“From his first week, Aurélien put in place strict typing conventions and a robust CI that cut our front-end regressions in half.”
Front-end (React) for Nabla Copilot, an LLM-powered clinical assistant. I owned the web application and the Chrome extension: token-by-token text streaming, asynchronous state shifting while the doctor is mid-consultation, and a perceived-latency budget close to zero. Building a dependable interface on top of a non-deterministic system is an architecture problem before it is a UI problem. Designed and shipped the web app's V1 in two weeks by building on the existing Chrome extension, with only Copilot autocomplete (no agents yet), then rolled out later iterations progressively behind feature flags.
Went from software engineering intern to front-end lead, with technical responsibility for critical healthcare and energy projects. Healthcare (bioMérieux): front-end architecture and full-stack development (Angular, Spring Boot) of medical applications handling sensitive data for millions of patients. Led the cloud migration to AWS (Lambda, Cognito, Terraform) and mentored junior developers. Energy (EDF): designed an offline-first Angular tablet application for field interventions, where state management without a connection and performance were the main challenges. Internal engineering: built Thales' technical assessment platform from scratch, with a secure, isolated code execution environment in Docker.