I am an AI Tech Lead building backend systems, AI agents, automation tools and full-stack products.

Some of my projects are serious proof-of-concepts around RAG, agents, dashboards and data pipelines. Some are deliberately fun, like small macOS utilities that turn system sensors into playful experiences. In both cases, I care about the same things: clean architecture, reliable APIs, practical UX and code that other people can understand, reuse or improve.

I like open projects because they make ideas easier to share. A lot of what I build starts as an experiment, then becomes a reusable tool, a demo, a package, a CLI or a small product.

• Tech lead in AI, with a strong focus on practical delivery, architecture and developer workflows
• Building AI agents, RAG systems, local agent runtimes and automation workflows
• Exploring LLMs, OpenAI/OpenRouter, LangChain, LangGraph, Qdrant and vector search
• Strong backend background with PHP, Laravel, Symfony, Node.js, TypeScript and API design
• Comfortable shipping full products with React, Next.js, Vue, Docker, databases and CI
• Often building things for fun, but with an expert engineering mindset
• Open to collaboration on AI tools, developer products, bots, SaaS platforms and weird useful experiments

• MacRaclette: native macOS menu bar thermal monitor that tells you when your Mac is hot enough to melt cheese. Funny idea, real SwiftUI app, live sensors, fan RPM, thermal stats and no telemetry.
• MacBonk: tiny macOS menu bar app that detects laptop impact spikes and plays custom sounds. Part toy, part hardware/sensor experiment, part configurable utility.
• Timeline: React interface for training/program timeline generation and export workflows.
• best-off-student: Docusaurus/PWA documentation and project showcase.
• burn-up-down: lightweight React experiment.

I also work on AI agents, RAG systems, automation workflows and data-heavy prototypes. Some of that work lives outside public GitHub, so I keep the public profile focused on the repositories that are actually visible here.

• AI agents with tools, memory, streaming responses and real application context
• RAG pipelines that turn messy documents into useful searchable knowledge
• Local-first developer tools, CLIs and automation dashboards
• Backend APIs, queues, auth, database models and deployment-ready services
• Small open-source apps that are fun enough to remember and solid enough to use
• Product prototypes that prove an idea quickly without ignoring engineering quality

• Email: [email protected]
• LinkedIn: linkedin.com/in/alois-marcellin
• GitHub: github.com/MagnusDot