// AI assistants for small teams & entrepreneurs
I build AI assistants that take admin off your plate, for small teams and entrepreneurs: email, spreadsheets, purchasing and reporting. They do the busywork, and you approve anything that matters. Behind them are 8+ years of building production software that has to run every day.
Assistants that triage email, track deadlines and funding calls, keep spreadsheets up to date, prepare purchases and write briefings. They live where you or your team already talk (Discord or a web chat) and connect to Google Sheets and the tools you use. Anything with an outside effect, like sending an email, placing an order or paying, waits for you to approve it.
Autonomous agents that run engineering operations: standups, pull requests and research. Custom tooling, isolated services and predictable costs.
Training non-technical teams to use AI on their daily tasks, from state-sponsored bootcamps to redesigning internal processes.
CI/CD with GitHub Actions and Jenkins, cloud on AWS and Azure, and full-stack delivery in TypeScript, serverless and Unity/XR. CTO at Xarp Reality Labs, President at coop.tech.
An assistant for the operations lead at coop.tech, a Portuguese tech cooperative. It triages her email, tracks funding calls, keeps her task sheet up to date and sends a daily briefing. It drafts, and she approves anything that goes out.
Read case study →coop.tech members ask for a purchase in chat. The assistant checks the project budget, fills the cart on the real shop website and reports back. A person reviews the cart and places the order.
Read case study →A self-hosted multi-agent system that runs daily engineering operations: one agent ships pull requests, one runs team standups end to end, one handles deep research. Isolated services, custom tooling, predictable costs.
Read case study →An AI education tool used in state-sponsored bootcamps, plus internal adoption programs that turn everyday tasks into AI-assisted workflows, built for people who don't write code.
Read case study →A Python framework I built from scratch for always-on Claude agents on Linux. Each agent is an isolated service with its own chat control surface, conversation memory, cost-aware triggers and a work ledger that logs every run. It adds a live dashboard, voice notes in and spoken replies out, a coordinator agent that delegates long tasks, and deploys itself when changes merge.
// private repository, details on requestFixed-scope packages for an AI assistant that works for your team.
Best for: one job you want off your plate.
Price on request
Order on Upwork ↗Best for: a team that wants the assistant shaped around how it works.
Price on request
Order on Upwork ↗Best for: several jobs, including work on websites with no API.
Price on request
Order on Upwork ↗Hosting checks, updates and small changes after launch. Sold separately.
Price on request
Not sure which fits? Book a consultation ↗
I design, build and run the cooperative's AI assistants, including the back-office assistant and the purchasing assistant, on my own agent framework.
Building Physical AI: augmented reality, smart mirrors, and behavioural analytics. Xarp is the company behind Dressify. I lead product engineering and applied research, and part of our team operations runs on autonomous AI agents.
Founded the first Portuguese tech cooperative, a startup consortium and technological innovation center. CTVC holds a seat on the EU AI Act Advisory Forum, where it provides technical opinions to the European Commission on AI. It is one of only two cooperatives in the Forum.
Five years on CryptoFights: Ascension, a multiplayer mobile game with about 10K downloads and 1K monthly active players, where I owned the CI/CD pipelines (GitHub Actions and Jenkins), the Azure serverless backend integration, and the combat and matchmaking systems.
I also build AI that works in the physical world: cameras, sensors and devices that see, listen and measure, and software that acts on what they pick up.
Smart mirrors and AR virtual try-on for fashion retail, with computer vision and behavioural analytics. As CTO I lead product engineering and applied research.
xarp.ai ↗
A network of smart buoys with underwater sensors for monitoring estuaries and ports. First pilot: six buoys in the Lima estuary with the port authority, September 2026, feeding a live public dashboard.
aquascan.pt ↗
A local-first voice assistant: an ESP32-S3 satellite in a 3D-printed enclosure I designed, with on-device wake word and local speech-to-text and text-to-speech on a Raspberry Pi hub. Only text queries leave the house.
// personal project, in progressI spent 8 years shipping multiplayer games. That is where I learned to build software that stays up under real load.
I'm João, a software engineer from Viana do Castelo, Portugal. I build reliable software and AI assistants that real teams use every day, including the agents that run parts of my own companies. Over the last decade I've also built multiplayer games, XR products and SaaS.
I co-founded Xarp Reality Labs (where I'm CTO) and coop.tech, the first Portuguese tech cooperative. I care about software that holds up in production, automation that saves real time, and teaching people to get real value out of AI.
I also write about software, AI, and building things. You can read my posts on Substack.
