Case studies

The context, the team, the problem, what I did and what came out of it. Companies are described by type. Names are shared after a first call.

AI-assisted code review and a breaking-change detector

Context: Industrial IoT scale-up, part of a global industrial group.
Team: Squad of up to 7 engineers across frontend, backend and cloud.
Stack: Claude CodeTypeScriptCI/CDPull-request checks

Problem

Backward compatibility mattered across a platform with many connected sites, and review time alone could not guarantee it.

What I did

Rolled out Claude Code for development.
Introduced AI-assisted code review.
Built a breaking-change detector integrated into pull requests.

Result

The detector runs on every pull request, so a breaking change is flagged in review.

Industrial IoT platform, from single instance to auto-scaling

Context: Industrial IoT scale-up. Roughly 600 sites and 2,000+ connected devices sending telemetry every minute, millions of readings per day.
Team: Squad of up to 7 engineers. I hired 4 of them.
Stack: Node.jsTypeScriptReactMQTTAWSKubernetesEvent-driven

Problem

Key services could only run as a single instance in production. That capped scale and made the platform fragile.

What I did

Drove the move to an event-driven architecture.
Solved the idempotency issues that blocked running several instances.
Set up architecture reviews and design docs.
Delivered the integration into the group's global cloud with engineering teams abroad.

Result

Services that were limited to one instance now auto-scale.

B2B payments and a paid subscription launch

Context: Fintech unicorn, B2B payments for small businesses.
Team: 4 engineers.
Stack: Node.jsReactAWSMySQLRedisMongoDB

Problem

First, an accounting integration: data sync and in-app invoice payment. Then a go-to-market program that needed sign-up, billing and cancellation across several teams.

What I did

Led the team that owned the accounting integration.
Moved with the team to the paid subscriptions program, owning the frontend.
Coordinated integration across the payment, subscription and app backend teams.
Owned technical design and code reviews.

Result

The integration shipped to small business customers and the team took the subscription frontend through the program.

Computer vision for retail shelves

Context: Retail computer vision company.
Team: Up to 8 engineers building APIs, web apps and Android and iOS integrations.
Stack: AWSGCPGraphQLKotlinSwift

Problem

Retailers needed to know what was missing or misplaced on the shelf, and store staff needed to act on it.

What I did

Delivered a mobile app for store management and shelf replenishment.
Built a solution on small in-store cameras that detects out-of-stock and misplaced products and raises alerts.
Added planogram compliance checks.
Owned solution design, version management and releases.

Result

Detection, alerts and the replenishment app shipped to retail customers.

A personal project built with an AI agent, and this site

Context: A personal project, and the site you are reading.
Team: Me and AI agents.
Stack: Claude APINestJSReactMySQL

Problem

Can one person build a real product with an agent doing part of the work? And can a personal site show that instead of claiming it?

What I did

Built an agent into the project: entity creation, analytics, support.
Built this site with AI agents.
Connected a daily agent that feeds the private side of this site through an API with scoped keys.
Put a filter on every public AI answer, with an automated leak test.

Result

The Ask my AI and Match your job features on this site run on this architecture. You can try them.