Selected work

Work

Proof, not promises. A data-first track record across energy and regulated finance, plus the range to design, build, and ship entire platforms and products single-handedly, from messy real-world constraints to systems that are live, monitored, and yours to run.

TrainSphere

A marketplace of certified sports coaches, designed, built, and hosted A→Z, solo.

TrainSphere is a two-sided marketplace connecting certified sports coaches with people looking for structured training. Thomas Baudry designed, built, and operates it end to end, from empty repository to a live, monitored, GDPR-compliant service. Coach trust is enforced by verification of credentials, business registration, and identity in the data layer.

Technology

Frontend
Next.js 15 (App Router), React, TypeScript strict
Backend & data
Raw SQL via pg on PostgreSQL (AWS RDS), 60+ versioned migrations
Infrastructure
ECS Fargate, ALB + HTTPS, private VPC, Terraform (eu-west-1)
Delivery
GitHub → CodeBuild → ECR → rolling ECS deploys
Payments
Stripe-ready payment layer (FR-compliant)

How it ships

RepositoryGitHubCI/CDCodeBuild → ECRECS FargateALB + HTTPS · RDSUserstrainsphere.fr
Every push flows from the repository through CI/CD to a rolling deployment on ECS Fargate. It reaches users in minutes, with a health check gating each release.

Case studies

Long-running, production-grade engagements in data-intensive, often regulated industries.

EDF Renouvelables

Renewable energy

AWS data engineering and architecture for solar parks. A long-running production engagement building reliable data systems in renewable energy, the backbone of operational reporting and analytics.

CFM / Capital Fund Management

Regulated quant finance

Data Platform Engineer in an audited environment: migrated on-premise Oracle to AWS and batch pipelines to Airflow, and standardized pipeline and architecture patterns. Enabled around thirty data analysts on data-engineering best practices, and took part in the company’s Claude Code community of practice, driving day-to-day adoption. Also built an AI reviewer for merge requests, adopted company-wide, that returns a summary, actionable comments, and a change verdict.

  • ~30data analysts enabled
  • ~500projects covered

Klarsen

AdTech / marketing data

Data engineering and architecture on AWS with end-to-end GDPR compliance. Measurable execution from pipeline design to production.

  • ×7execution velocity
  • <1%defect rate
  • 100%repeatability

Crédit Foncier

Finance / credit

Production management and C#/SSIS development for credit business intelligence. Operational reliability in a finance environment.

Other projects

A wider range: data platforms, AI apps, and automation. Each one shipped, not slideware.

MantaAI

Turns a LinkedIn export into business intelligence: multi-source enrichment, AI scoring, and a relationship graph (PostgreSQL, Neo4j, ChromaDB).

linkedin-network-analyzer

The same idea, fully local and free: privacy-first, on Ollama and pgvector with no costly cloud APIs.

ai-orchestration

A security-first multi-agent framework: 18+ specialized agents, 133 tests.

trainsphere-marketing

A social-content engine for TrainSphere: marketing-ops and cost engineering, with human-in-the-loop publishing.

Let’s talk.

Data platforms, full products, or AI enablement, as staff augmentation or a scoped deliverable. Tell me what you need running in production.

Get in touch