Projects that show systems thinking in practice

A business-facing booking product needs to handle guests, resources, times, staff roles, and conflicts without creating friction for the user.

I built a Laravel platform with public booking without accounts, admin and staff views, RBAC, tenant isolation, and transaction-safe booking conflict checks.

Shows that I can build systems where business rules, data model, and daily workflows need to hold together, not just a polished frontend shell.

Exchange rates need comparison over time with KPIs, charts, and risk indicators without heavy backend operations.

I built a monorepo with a React/TypeScript frontend and Python pipeline that fetches ECB data, publishes static JSON files, and updates daily through GitHub Actions.

Shows that I can connect data, CI, performance, and interactive UX into a solution that is easy to operate and reason about.

Visitors need to understand the course, setup, and results quickly, while the organizers need a secure way to register players, rounds, and courses without touching the database directly.

I built a React/Vite frontend that talks to a purpose-built ASP.NET Core backend: public season and player pages, an admin panel behind JWT login for managing players, rounds, and courses, and an adapter layer that normalizes the API's varying response shapes.

Shows that I can build and wire a frontend to a self-authored authenticated backend across two repos, with adapters and a resilient API client that keep the UI decoupled from the backend's exact response shape.

Tracking a round on paper gives you no data to improve from. Golfers need a way to log rounds hole by hole and get concrete feedback on what actually cost strokes.

I built a full-stack Next.js app with a hole-by-hole scorecard, localStorage persistence, statistics charts, screenshot import via AI vision, and Claude-powered coaching analysis per round.

Shows end-to-end product thinking: secure server-side AI integration, structured data from unstructured input, and a UI that stays usable during a round.

Local football coverage is scattered across sources and local newspapers. No single place aggregates standings, matches, and goal scorers for a town's teams, kept automatically up to date.

I built a Next.js app that pulls leagues, matches, and standings from an external sports data source, and lets Claude read local newspaper match reports to judge relevance and extract goal scorers from unstructured text.

Shows that I can build a source-agnostic ingest architecture and apply AI to structured extraction from text, running on a scheduled, cost-free operational setup.

Splitting expenses in a group (a trip, a shared cabin, a dinner club) quickly turns into a tangle of small debts crossing between everyone, and nobody's quite sure who owes whom.

I built a Laravel + Inertia + Vue app with groups, split expenses, and a greedy debt-simplification algorithm that computes the minimum number of transactions needed to settle the whole group, plus reminders with an atomically-protected cooldown.

Shows I can solve a real algorithmic problem, not just build CRUD, and handle race conditions correctly with atomic database operations.

Turning features on and off in production without control over who approves it, traceability for what changed, or a fast rollback path is dangerous once multiple environments and multiple people are involved.

I built a Laravel-based feature-flag control plane in the style of LaunchDarkly: project-scoped API keys, deterministic rollout hashing, a draft/publish workflow per environment with two-person approval for critical flags, a break-glass emergency override, and an immutable audit log with a checksum chain.

Shows I can design governance and permissions for a tool other developers trust in production, not just build features — governance, traceability, and race-condition safety are first-class citizens, not afterthoughts.