I'm a software engineer with 6 years of professional experience building production systems, primarily with C#, .NET and Azure, with a strong focus on backend engineering, distributed systems, performance and architecture.
More recently, I've been working at the intersection of software engineering and AI, particularly interested in how autonomous agents can be incorporated into real engineering workflows without sacrificing control, reliability or safety.
Engineering is about solving problems. Well-crafted code is the consequence, not the goal.
I'm fundamentally tool-agnostic: the problem should dictate the architecture, technology and level of abstraction.
My current focus is AI engineering rather than simply AI integration.
I'm particularly interested in:
- 🤖 Agentic systems & autonomous software engineering
- 🧩 Multi-agent orchestration and agent coordination
- 🏗️ System architecture and bounded execution
- 🔒 Sandboxing, isolation and security boundaries
- 🔄 State machines and deterministic orchestration
- 📋 Product → feature → story → task decomposition
- 🛠️ AI-powered developer tooling
- 📊 Observability, cost and execution tracking
- 🔌 Provider-agnostic AI integrations
- ⚡ Performance, scalability and reliability
I treat AI agents as engineering components inside a larger system, rather than treating an LLM as the system itself.
VibeBoard is a local-first cockpit for building software with AI coding agents.
The core idea is simple:
Give agents enough freedom to do meaningful engineering work, while keeping the surrounding system in control.
The project combines a product-oriented Kanban workflow with autonomous coding agents. Work is represented through three linked levels:
Features → Stories → Tasks
Agents can plan, implement, review and test work while the system maintains the state and boundaries around their execution.
VibeBoard deliberately separates decision-making, persistence, orchestration and execution.
- 🧠 Deterministic core — orchestration decisions are driven by a state machine rather than prompts
- 🗂️ Markdown-based project state — cards are files, making project state readable, versionable and Git-friendly
- ⚙️ Agent orchestration — autonomous workflows coordinate planning, implementation, review and testing
- 🐳 Containerised execution — each project gets isolated agent containers
- 🔐 Capability boundaries — agents operate within controlled filesystem and network boundaries
- 🔑 Per-card credentials — agents interact with the project through constrained API access
- 📡 Event-driven updates — filesystem changes are propagated live between the board and agents
- 📊 Execution observability — runs track time, tokens and cost
- 🛑 Operational controls — execution can be stopped at multiple levels, including terminating all agents
The architecture is intentionally designed around controlled autonomy: the agent can be powerful inside its execution boundary without becoming a trusted component of the host system.
My professional work is primarily focused on backend and enterprise systems.
I've worked with:
- High-volume and data-intensive workflows
- .NET microservices
- Distributed systems
- Azure infrastructure
- SQL performance optimisation
- CI/CD and deployment automation
- Enterprise financial systems
- Full-stack product development
- LLM-powered internal applications
One example from production work: I redesigned a high-volume reconciliation process and reduced execution time from 45 seconds to 1.3 seconds through database indexing and query redesign.
A multi-tenant platform for real-time anonymous Q&A.
Built with React, Python, Flask, PostgreSQL and Server-Sent Events, with real-time synchronisation, voting, queue management and host controls.
A full-stack application built with SvelteKit and TypeScript, integrating a self-hosted Mistral-7B model for AI-assisted content generation.
The project explores practical LLM integration while maintaining control over the application architecture and model deployment.
LLM engineering · AI agents · RAG · MCP servers · self-hosted models
TDD · mutation testing · component-driven development · Agile
ATHE Level 7 Diploma in Computing Technologies — Artificial Intelligence
Continuing to deepen my knowledge across AI, software engineering and the systems required to turn AI capabilities into reliable software.
🇧🇷 Portuguese — Native
🇬🇧 English — Fluent
🇮🇹 Italian — Advanced
🇪🇸 Spanish — Intermediate