Software Engineer · Paddle

Gonçalo Raposo

Software engineer with a background in aerospace engineering and artificial intelligence research.

Currently working on backend systems at Paddle, primarily with Go and Python.

Gonçalo Raposo

About

My path runs from Aerospace Engineering through AI and NLP research to production software. I studied at Instituto Superior Técnico, researched at INESC-ID, worked at Unbabel, and am now a Software Engineer at Paddle.

Full professional experience on LinkedIn

Selected work

Notícias do Dia public interface

Notícias do Dia

A daily European Portuguese news podcast built from curated RSS feeds, automated with GitHub Actions and published through RSS.

Separates news collection, audio generation and RSS publishing into repeatable workflow steps.

  • Python
  • Automation
  • RSS
  • Audio
Workout Player public interface

Workout Player

A web app for running AI-generated workouts, with timers, circuits, progress tracking, wake lock and sound cues.

Turns a structured workout into timed steps, including rest, repetitions and sound cues.

  • TypeScript
  • Web
  • PWA
Orbit Radar · Data → orbits → visualisation

Orbit Radar

A satellite tracker with 3D visualisation, live positions, ground tracks, search and camera follow.

Combines orbital data and propagation with an interactive view of the Earth.

  • React
  • TypeScript
  • Orbit
Split public interface

Split

A receipt-splitting app with capture, perspective correction, item extraction, editing and totals.

Keeps receipt extraction editable before assigning items and calculating totals.

  • React
  • TypeScript
  • Gemini
PROJECT OVERVIEWAI agentMCPScilab / XcosInspect · build · validate · simulate

xcos-mcp

An MCP server that lets AI agents create, modify, validate and simulate Scilab/Xcos models.

Connects an agent’s tool calls to model validation and simulation in an existing engineering environment.

  • Python
  • MCP
  • Scilab/Xcos
  • Simulation
PROJECT OVERVIEWNeural networkMixed precisionPositNNC++ · mixed low-precision arithmetic

PositNN

Master’s thesis · HPC-ULisboa repository

A C++ framework for training and inference with neural networks using low-precision posit arithmetic.

Explores mixed low-precision arithmetic as part of my master’s thesis.

  • C++
  • Posits
  • Neural networks
  • Systems

Latest writing

Research

PositNN: Training Deep Neural Networks with Mixed Low-Precision Posit

My master’s thesis work on training deep neural networks with mixed low-precision posit arithmetic.

Gonçalo Raposo, Pedro Tomás, Nuno Roma · ICASSP · 2021

Information Retrieval Using Fuzzy Fingerprints

Exploring fuzzy fingerprints for information retrieval.

Gonçalo Raposo, João Paulo Carvalho, Luísa Coheur, Bruno Martins · IPMU · 2024

Prompting, Retrieval, Training: An exploration of different approaches for task-oriented dialogue generation

Comparing prompting, retrieval and training for task-oriented dialogue generation.

Gonçalo Raposo, Luísa Coheur, Bruno Martins · SIGDIAL · 2023

Question Rewriting? Assessing Its Importance for Conversational Question Answering

Studying question rewriting in conversational QA, linked to the first-place SCAI QReCC 2021 system.

Gonçalo Raposo, Rui Ribeiro, Bruno Martins, Luísa Coheur · ECIR · 2022

Document-Level Abstractive Summarization

An exploration of abstractive summarisation at document level.

Gonçalo Raposo, Afonso Raposo, Ana Sofia Carmo · Instituto Superior Técnico · 2022

Beyond code

Outside software, I spend a fair amount of time running, travelling and falling into new technical rabbit holes. I completed an Ironman in 2024 and still enjoy building and trying new things.

Contact

For software, research or project conversations, find me on LinkedIn or GitHub.