• Solely owned a new Azure backend platform end-to-end (architecture, infrastructure, deployment, monitoring, and maintenance) for Swoove's pivot into generative AI, integrated with the existing backend for authentication and animation data.
• Built Python and C# serverless Function Apps for LLM-based script creation and editing, content-aware animation selection, and ElevenLabs text-to-speech orchestration.
• Developed and operated REST APIs serving an international user base distributed primarily across the US and Europe.
• Designed Redis-backed concurrency controls that enforced provider rate limits across scaled instances by tracking in-flight voice-generation jobs and releasing capacity on completion or timeout.

See below for more

• Used PostgreSQL and Cosmos DB for production data, preprocessed animation metadata into Redis for low-latency access, and implemented LLM prompt caching to cut API cost and latency.
• Covered the services with pytest and NUnit unit and integration tests, used DSPy for structured generation and error checking of the LLMs.
• Instrumented structured logging, request/error/latency telemetry, dashboards, and alerts in Application Insights.
• Trained a humanoid pose-prediction model and deployed it for on-device inference via Barracuda in Unity; also contributed client-side features to the Unity app.
• Automated Function App deployments and CI/CD with Azure Pipelines, later migrating infrastructure provisioning to Terraform.

Python, C#, FastAPI, Flask, FastEndpoints, DSPy, PostgreSQL, Redis, Cosmos DB, Azure Functions, Azure Pipelines, Terraform, Application Insights, ElevenLabs, Unity

• Co-developed EdAR, a Unity-based educational AR demonstrator. Owned its Python/AWS backend and co-owned Unity rendering and experience integration.
• Built a serverless content pipeline using AWS Lambda, S3, Aurora, and JWT authentication; parameterised scene descriptors let creators publish or update AR experiences without rebuilding the app.
• Developed an industrial quality-control demonstrator in C++ and OpenGL, plus full-stack Mixed Reality demonstrations for HoloLens 2.
• Represented the department in client visits, sales meetings, conferences, and media initiatives.

Python, AWS Lambda, S3, Aurora, JWT, Unity, C++, OpenGL, C#, HoloLens

See below for major projects

The Minho Advanced Computing Center (MACC) is a sustainable supercomputing and bigdata infrastructure catering to national scientific and industrial communities and complementary to international partners. Due to the Covid-19 pandemic, the project encountered delays in construction and furnishing of the building set to host the Deucalion supercomputer, so this VR tour was created to present MACC at the 2021 High Performance Computing for Science and Innovation event.

See more: #1 #2

The EdAR project consists of an extensible, Unity-based, framework for the development of custom AR experiences for education for phones and tablets. The project was created with the goal of making AR content creation accessible to academic staff, providing an easy way for teachers to augment their teaching with multimedia content in AR. In partnership with the University of Edinburgh, I co-developed the engine that powered the experiences, implemented several, and was the developer responsible for the networking implementation, which enabled the application to stream content from AWS, keeping content updated with the least amount of possible friction for creators.

See more

• Sole architect and developer of an EU-funded real-time Mixed Reality remote assistance system for industrial maintenance, delivered end-to-end from co-authoring the funding proposal to final validation by the industrial partner and funding body.
• Built the full stack: a Nest.js and MongoDB backend (user management, asset CRUD with an extensible storage abstraction, WebSocket session brokering), an Angular frontend, and a HoloLens app, with WebRTC peer-to-peer audio/video and WebSocket streaming of real-time AR annotations, spatial coordinates, and signalling.
• Worked around the immature HoloLens streaming ecosystem with low-level capture approaches to stream video enriched with spatial mapping data, a use case with little official or third-party support at the time.
• Drove the real-time call and spatial annotation features through end-to-end multi-device scenarios across HoloLens and browser clients, covering session brokering, signalling, and reconnection paths.

See below for more

• Maintained the department's shared server environment: Dockerised and deployed research services, and helped colleagues containerise their own projects.
• Developed full-stack AR/VR/MR experiences and their backend/cloud architecture (AWS services, real-time streaming, content delivery) across multiple EU and client-facing projects, including AR apps driven by dynamic content from remote CRM/CMS systems; coached new colleagues in AR/VR/MR development.
• Represented the department in client visits, sales meetings, and conferences; assisted with project planning and budgeting.

Node.js, Nest.js, TypeScript, JavaScript, MongoDB, WebRTC, WebSockets, Angular, Python, AWS, Docker, Unity, C#, HoloLens, Android

See below for major projects

The Bosch CMMS (Computerized Maintenance Management System) project aimed to develop an innovative new system for management of scheduled maintenance interventions. The proposed system resulted in an upgrade of the existing system through a new architecture and a set of advanced management methodologies and functionalities. In particular, my role was to implement a new mixed reality application using Microsoft's Hololens, capable of presenting, both in real-time and in situ, a plethora of information available in the existing systems and data stores, such as documentation and maintenance data.

The Enhancing Shopping Experience project is a multifaceted project, with one side focused on providing retailers with a toolset to better engage costumers, by offering a set of interactive, immerse, and innovating experiences, both within the shop and without. Where consumers are concerned, this project aims to bridge the online and offline retail channels into a more unified experience, allowing users a more seamless and informed purchasing process. My role in this project was the design of the architecture and development of an extensible AR application, consuming custom scene descriptors from a remote CRM, to present a dynamic and engaging experience, with a particular focus on the presentation of, and interaction with, custom multimedia content.

This project aims for the exploration of the tactile sensation in the design of specific components integrated into cars interiors, by developing a tactile probe that extracts features from real manufactured parts, coupled with a computational model to correlate the gathered data with the features in the part. To complement this, a virtual prototyping and visualisation system is to be integrated into the QC process, to provide visual and tactile feedback before the mould is produced. My contribution to this project, which was still in its early stages, was in the development of this latter virtual system, using an HTC Vive and VR gloves to allow the user to be immersed and interact with the pre-manufacture prototype.

Development of a software service-oriented architecture and technology solutions, under the paradigm of IoT and Industry 4.0, which promotes a corporate and aggregate vision of industrial units operations dispersed across different geographies through remote and local access; building collaborative and cross-cutting tools; optimisation of operations and user experience in industrial units and, finally, system reliability. My main task focused on the development of a full-stack solution for Remote Assistance using Microsoft's Hololens on one end, and web technologies such as Angular and Node.JS on the other, featuring real-time audio/video communications, bidirectional annotations in the physical environment, and user-defined multimedia content.