vasconunes

Vasco Nunes

Backend Developer

I’m Vasco, a backend developer who enjoys building applications inspired by everyday experiences. I mainly work with Java and Spring Boot, and explore AWS and AI integration through my projects.

Vasco Nunes

About

I studied Computer Science and Engineering at NOVA FCT in Lisbon, where I completed my bachelor’s degree and recently submitted my master’s thesis. I also spent an exchange semester at Universidad Autónoma de Madrid.

I worked on my thesis during a research internship at ARDITI, exploring how large language models could help people model business processes and data in a low-code platform.

I’m currently building AnamVis, an app that helps patients and doctors communicate, inspired by my father’s experience as an ophthalmologist.

Outside of software, I enjoy football and kickboxing, and I go to the beach for a swim whenever I can.

Projects

AnamVis

–Present

Patient communication platform

My father is an ophthalmologist, and seeing his WhatsApp fill up with patient messages inspired me to build AnamVis. It brings patient messages and medical records into one place, with AI-assisted triage to help doctors identify which cases may need attention first.

  • Structured AI-assisted triage.
  • Patient history search with verifiable sources.

Java, Spring Boot, Spring AI, PostgreSQL, AWS S3

Patient follow-up
1 / 6

IronBuddy

-

Training with friends makes the gym more motivating and social for me. Joining a running club to meet people during a summer school in Nice helped shape IronBuddy: a community for finding training partners, meeting your gym crush and staying active together.

Java, Spring Boot, PostgreSQL, React, WebSockets, Cloudinary

Login
1 / 8

DISME Wizard

LLM-assisted business modelling
Master’s thesis, NOVA University Lisbon

Turning business requirements into a low-code application requires modelling knowledge. I built a wizard that guides users through natural-language interviews and generates models, queries and business rules for DISME.

Users can review and refine the generated models before importing them into DISME, where they become forms and executable workflows.

Python, Streamlit, Ollama, JSON Schema, Blockly

Read thesis
Requirements interview
1 / 4

Experience

Research Intern

ARDITI Remote, Funchal, Madeira

–

Research internship carried out as part of my master's thesis at NOVA FCT, focused on LLM-assisted business modelling for the DISME platform.

Software Development Intern

Cult of Bits Lisbon, Portugal

Developed React and TypeScript features for an internal dashboard and integrated them with REST APIs.

Worked with the development team on debugging, refactoring and UI performance improvements.

Research

Large Language Models in Low-Code and Model-Driven Platforms: A Systematic Literature Review

Systematic Literature Review

A review of 35 studies examining LLM integration, intermediate representations and validation in low-code and model-driven platforms.

Read review

Courses and Summer Schools

Cybersecurity Summer School

Algebra Bernays University Zagreb, Croatia

View certificate

Generative Artificial Intelligence for Industry

CESI Nice, France

View certificate

Generative AI with Large Language Models

DeepLearning.AI

View certificate

Contact

Feel free to get in touch.

vasconunes200@gmail.com LinkedIn
GitHub activity

Loading contributions…

LessMore

CV

Download PDF
Vasco Nunes CV

AnamVis

My father is an ophthalmologist. When we’re spending time together, like having dinner or watching a football match, he often receives messages and calls from patients. I’d see his WhatsApp full of patient messages that he’d reply to whenever he could. It made me think about how difficult it can be to give each patient the attention they need while keeping up with everyday life.

I started AnamVis to bring these conversations and each patient’s medical information into one place, including test results, documents and updates after surgery. Patients can describe their concerns, and doctors can review them alongside their history, without having to piece together past conversations or rely on memory.

IronBuddy

I enjoy going to the gym, and training with friends makes it more motivating and social for me. I wanted to make it easier to find that company, whether at home in Lisbon or somewhere new.

During an AI summer school in Nice, I joined a running club to meet people. That experience helped shape IronBuddy: a community where people can meet others, find training partners and stay active together.

There’s a feed for sharing workout photos, with likes and comments from friends, and a forum for organising runs, discussing training, sharing healthy recipes or asking for local recommendations, like where to train when visiting a new city. You can also see which friends are currently at the gym, making it easier to join them for a workout.

Conversations with friends about people they’d seen at the gym and found attractive, but hadn’t felt confident enough to approach, also led to Gym Crush, a feature that helps people connect when there’s mutual interest.

DISME Wizard

Generation of Business Models in a Low-Code Platform with Large Language Models

NOVA University Lisbon
MSc in Computer Science and Engineering

For my master’s thesis, I explored how large language models could help users turn business requirements into executable applications in DISME, a low-code platform based on Enterprise Engineering. I built a Python and Streamlit wizard that guides users through natural-language interviews, generates models, queries and action rules, and lets users review them before importing them into the platform.

The main challenges were keeping the LLM’s outputs within DISME’s modelling constraints and making the interview questions clear and relevant to what users had already explained. Integrating the generated rules with the existing Blockly functionality also required adapting and validating how they were translated into blocks that the platform could execute.

In an accompanied study with 19 participants, all completed the assessed wizard tasks and 17 completed the tested execution path in DISME. Every participant needed semantic assistance, highlighting where the questions, explanations and review process still needed improvement.

Read full thesis

Project screenshot