I am a Full Stack Developer with a practical mindset and a passion for building scalable solutions that solve real-world problems. I specialize in modern web ecosystems and business automation, focusing on creating efficient and maintainable products.

• Currently: Junior Full Stack Developer at AR IT Services.
• Education: Studying Information Systems Engineering at UTN-FRC and Software Development at Inst. Brochero (100% Merit-based Scholarship). Full Stack Developer Career at Coderhouse

Event-driven workflow engine on directed acyclic graphs (DAGs).

• The Stack: Python, FastAPI, PostgreSQL, asyncio, React, TypeScript, React Flow.
• The Solution: Async parallel DAG execution, JSONPath expressions for dynamic payloads, and Time-Travel Debugging via event-sourcing (replay the historical state of any run, node by node).
• Extras: Visual graph editor built with React Flow.

Multi-store ecommerce builder, SaaS-style (à la Shopify / Tienda Nube).

• The Stack: Go, MongoDB, Redis, SvelteKit, TypeScript, Docker, microservices.
• The Solution: Never oversell stock under concurrent load, isolated per store — atomic MongoDB operations and time-boxed cart holds with Redis, in a multi-tenant Go microservices architecture.
• Extras: SvelteKit frontend covering marketing, storefront per store, and merchant dashboard.

REST API for membership management at a health/sports club.

• The Stack: Java 21, Spring Boot 3.3, MySQL 8, Flyway, MapStruct, Resilience4j, Bucket4j, Docker, GitHub Actions.
• The Solution: On-demand membership state engine (no caching). Real concurrency control: an initial @Transactional + SELECT FOR UPDATE attempt failed under load (commit happens after method return) — fixed with a per-client JVM lock + TransactionTemplate, verified with a 10-parallel-request test against a quota of 1.
• Extras: 27 tests (H2 for unit/controller, Testcontainers for integration), retry + circuit breaker on outgoing webhooks, rate limiting, and plan-catalog caching. CI/CD with GitHub Actions.

Chaos Engineering playground with real OTP processes, not simulations.

• The Stack: Elixir, Phoenix LiveView, PostgreSQL, Tailwind, Docker.
• The Solution: Every infrastructure node on the canvas (load balancer, API, DB, cache, queue) is a real GenServer under a DynamicSupervisor — "killing" a node runs an actual Process.exit(pid, :kill). Deliberate :temporary restart strategy to show real "let it crash" behavior, no auto-heal masking the failure.
• Extras: Live metrics dashboard (RPS, p99, error rate), automatic Chaos Monkey, and a centralized traffic simulator generating concurrent load. 101 tests running in CI/CD on every push to main.

Collaborative IDE for learning to code, with sandboxed execution of untrusted code.

• The Stack: .NET 8 (Clean Architecture + CQRS/MediatR), Angular 20, Firebase (Firestore, Realtime DB, Auth), Monaco Editor, Docker, Playwright.
• The Solution: A room holds at most 4 people, and the first optimistic Count < 4 check let two simultaneous joins both take the last seat. An integration test with 2 parallel joins against the real Firestore emulator caught it; fixed with a Firestore transaction (RunTransactionAsync).
• Extras: User code runs in ephemeral Docker containers (7 languages, no network, 256MB, 5s SIGKILL timeout, read-only filesystem, non-root), HTML/CSS graded with headless Chromium. Real-time Monaco sync with remote cursors and chat, AI Coach (OpenRouter) with rate limiting and a fallback to pre-generated hints, 34 seeded challenges, 58 unit tests plus Playwright E2E.

• Email: mateopavoni6@gmail.com
• LinkedIn: mateopavonidev
• GitHub: mateopavoni