Damien Gilbrin

Technical optimization and evolution of business tools for SNCF

For SNCF Gares & Connexions, I carried out a dual mission: resolving critical backend performance issues and evolving the internal tool used daily by station agents.

1. Performance Crisis Resolution Faced with saturated APIs impacting passenger information (response times > 10s, 50% errors), I deployed a "surgical" optimization strategy based on data analysis (Datadog):

Actions: Creation of partial indexes on PostgreSQL and refactoring of Lambdas (Node.js/TypeScript) to switch to bulk inserts.

Results: I reduced response times tenfold (1.1s to 102ms), brought the error rate back to 0%, and reduced server load by 74%.

2. Evolution of the tool for Passenger Information Operators (OIV) My intervention was not limited to code: I supported the product's evolution. I was proactive in improving the internal tool for OIVs, the agents responsible for announcements and display in stations. I designed and implemented several functional enhancements to streamline their management interface and secure their real-time operations.

Outcome: This mission restored total technical reliability while adding business value for the operational teams. I also delivered a long-term vision, recommending an "Access Pattern" architecture and a migration to DynamoDB to support future load.

- Implementation of real-time messaging between Skiller and its clients (using WebSocket technology with Pusher)

- Creation of the new live video conferencing system for up to 250 people (using AWS Chime SDK technology)

- Implementation of the automated Live Replay system (via AWS MediaPipeline, AWS EventBridge, AWS MediaConvert, AWS S3, AWS CloudFront)

- Creation of the "E-Learning" online training tool with video uploads, MCQ creation, free page creation (with Text, Image, formatting, PDF…)

- Implementation of an interactive calendar

- Implementation of authentication via Google, Apple, Facebook (via AWS Cognito) while migrating old accounts on the fly

- Study and initial implementation of artificial intelligence to generate offers in one click with automated descriptions, category selection, and AI-generated image creation (with AWS Bedrock, AWS StepFunction, AWS S3, and AWS Lambda)

- Performance optimization study (with AWS DynamoDb)

- Rapid development of a POC with VuetifyJs in Vue2

- Development of a new International version with Quasar and Vue 3 (Composition API)

- Direct front-end upload of very large files with transfer resume (via S3) to significantly relieve the API and storage server

- Study of external image miniaturization solutions (Cloudinary…) and POC of an internet solution with very low development/maintenance costs (via AWS CloudFront, AWS Lambda Edge, and S3)