Research engineer at INESC-ID building permissioned blockchain platforms for supply chain traceability. Previously developed robotics and computer-vision applications for healthcare at ISR-Lisboa.
Fullstack development and infrastructure leveraging Hyperledger Fabric blockchain technology for supply chain transparency and traceability, serving major Portuguese retailers.
A chaincode extension for Hyperledger Fabric enabling controlled mutability in permissioned blockchains that allows for time-limited data correction while preserving immutability and security.
Trained and deployed a YOLO computer vision model on the TEMI autonomous robot for real-time object detection and gesture recognition in hospital environments.
Built an autonomous navigation system with configurable waypoints and a touchscreen quiz interface for patient feedback collection during hospital rounds.
Developed a P2P video calling application using WebRTC on the TEMI robot, enabling remote patient visits and live Q&A sessions between patients and healthcare staff.
A modular adapter architecture that facilitates enterprise blockchain adoption by bridging legacy systems with decentralized networks through five key components, including data extractors, transformers, and messaging middleware. Successfully validated in a real-world supply chain pilot, the system enables small and medium-sized enterprises to achieve seamless interoperability and operational transparency with minimal disruption to existing workflows.
A traceability application leveraging Hyperledger Fabric to provide a decentralized "single source of truth", achieving optimal throughput and consistent latencies using 49 MB multi-transaction blocks. Tested across 10 independent peer instances, the system optimizes data sharing for multiple stakeholders while providing actionable insights into storage efficiency and memory usage for sustainable agro-food scalability.
EvoChain, a smart contract development framework, introduces controlled mutability to Hyperledger Fabric, enabling time-limited data corrections via a "grace period" while preserving long-term auditability. Validated through the an example supply chain based application, it offers operational flexibility with a minimal performance overhead of 9% throughput and 2% memory usage.
An intelligent self-hosted notification system that uses an UCB-Day algorithm UCB-Day algorithm to optimize health questionnaire timing by learning individual receptivity patterns. In a real-world pilot, the system achieved a 0.819 goal achievement rate with only 1.89 daily notifications.