Skip to content
View Luismpso's full-sized avatar
🦖
🦖

Highlights

  • Pro

Block or report Luismpso

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Luismpso/README.md

header banner

Typing SVG

LinkedIn Kaggle Email Based in Portugal profile views


👨‍💻 About me

  • 🎓 Master's in Artificial Intelligence @ Universidade do Minho · (2025 – present)
  • 📐 B.Sc. in Engineering Physics, Universidade do Minho — final grade 17/20 (ECTS A)
  • 🧭 Focused on designing, orchestrating and validating robust, scalable AI pipelines end to end
  • 📄 Corresponding author of a paper accepted at INISTA 2026
  • 🌍 Languages: Portuguese (native) · English (B2)
  • 📫 Reach me at [email protected]

📄 Publications

Safety-First Multi-Agent Chest-Pain Triage: A Manchester-Aligned Architecture that Surfaces LLM Under-Triage — Accepted at INISTA 2026

Multi-agent LLM system for emergency-department chest-pain triage, evaluated against real clinical data. Built on the Manchester Triage Protocol with a safety-first escalation design combining RAG, local LLMs and ML risk models. Developed with clinical collaborators from ULSAAve.


💼 Experience

📡 Machine Learning Intern — NOS · Jul 2026 – Sep 2026

  • Built an end-to-end data pipeline on BigQuery processing ~125M daily records to detect and rank mobile-network quality hotspots at national scale
  • Designed a two-level geospatial clustering system, re-architecting it to run inside BigQuery for scale (per-site runtime cut by an order of magnitude)
  • Developed a mobility-analysis module (community detection) and validated the model against 5,600+ customer-satisfaction survey responses
  • Delivered results through interactive Power BI dashboards on live data, with end-to-end system documentation
  • Stack: Python SQL BigQuery GCP Power BI geospatial analysis clustering

🐦 EMG Signal Acquisition Intern — BiRDLAB · Jun 2023 – Aug 2023

  • Built a real-time EMG acquisition & transmission pipeline on Arduino + STM32
  • Designed a 12th-order Butterworth band-pass filter (20–450 Hz) in C++ at a 1200 Hz sampling rate
  • Cut UART payload size by 75% with a custom float→16-bit encoding (stable 249,600 bps)
  • Stack: C++ STM32 Arduino UART Digital Signal Processing

🎓 Event Logistics Lead — NEFUM (Physics Students' Union, U. Minho) · Jul 2022 – Jul 2024

  • Directed logistics for 2 major academic events (~120 attendees each)
  • Planned and ran lectures & workshops for Physics / Engineering Physics students

🚀 Personal projects

Project What makes it interesting Stack
♟️ Grand Master AI Chess agent built from scratch — Dueling Double DQN with legal-action masking, Prioritized Experience Replay, potential-based reward shaping and a three-stage curriculum · Gymnasium-compatible env · 70 pytest tests Reinforcement Learning PyTorch Gymnasium

🏆 Featured academic work

Project What makes it interesting Stack
🥇 AI-Generated Text Detection 1st place in all 3 submissions (out of 25 teams) — multi-class classifier for Google / Mistral / Meta / OpenAI text · Grade 19/20 Deep Learning PyTorch
🩻 ERCP Medical Image Classification Multi-class deep learning with equipment-bias removal via SAM3 and model ensembles · Grade 19/20 Deep Learning PyTorch
🤖 FS Rules AI Assistant 100% offline RAG agent for Formula Student regulations, with accurate cited sources · Grade 18/20 Ollama Gemma 3 12B Flask
🏛️ SmartStudyRoom — Multi-Agent System Distributed library management with SPADE/XMPP, FIPA-ACL & Contract Net negotiation, real-time dashboard · Grade 20/20 Multi-Agent Systems Python
🏎️ Autonomous FS Racing Agent RL agent (SAC/PPO) in a 2D simulator with FS-AI rules, procedural tracks & domain randomization · Grade 19.5/20 Reinforcement Learning
🏅 Porto Traffic Flow Prediction 4th / 57 teams in a university Kaggle competition — full CRISP-DM pipeline · Grade 18.2/20 scikit-learn Pandas
📖 Study-Room Occupancy & Comfort Monitoring Low-cost distributed IoT system — 2× ESP32 nodes, YOLOv11 occupancy detection, web dashboard & Android app · Grade 19/20 ESP32 YOLOv11 IoT
🩺 Chest-Pain Triage System Manchester Protocol triage (Yellow–Blue) combining a multi-agent architecture, local LLMs, RAG & ML — basis for the INISTA 2026 paper · Grade 20/20 Multi-Agent Systems RAG LLM
⚡ D3.js Data Story — Energy Transition Reveals the two sides of the energy transition: renewables booming in electricity, fossil fuels still dominating primary energy · Grade 19/20 D3.js JavaScript
🛢️ D3.js Data Story — 2026 Hormuz Crisis Traces the price of the 2026 Iran war from the Strait of Hormuz to the Portuguese gas pump · Grade 20/20 D3.js JavaScript
🎯 Reinforcement Learning Portfolio 5 environments · 20+ algorithms (DP, MC, TD, SARSA, Q-Learning, REINFORCE, MCTS, DQN, AlphaZero) · 71 pytest tests · Grade 20/20 Gymnasium PyTorch
🎮 Millionaire — Prolog Edition Logic-driven game engine + GenAI "Phone a Friend" lifeline · Grade 17/20 Prolog LLM
🗺️ Multimodal Route Optimization (Porto) Bi-criterion shortest path (time + CO₂) over a complex transport graph · Grade 18/20 Nature-Inspired Computing
🏢 Tech Retail Decision Support System Star-constellation Data Warehouse (PostgreSQL, CSV, MongoDB) with NLP sentiment analysis & Metabase dashboards · Grade 17/20 PostgreSQL MongoDB NLP

Most academic projects above were built in teams of 3–4 — see each repository's commit history and contributor graph for individual contributions.


🛠️ Tech stack

Focus

AI & Machine Learning

Data Science & Analysis

Programming Languages

Web, DBs & DevOps

Hardware & Signals

footer

Pinned Loading

  1. AP1 AP1 Public

    🤖 Multi-class detection of AI-generated vs. human text in English, covering Google, Mistral, Meta, and OpenAI models.

    Jupyter Notebook 3

  2. AP2 AP2 Public

    🩻 Deep Learning models for multi-class ERCP classification featuring equipment bias removal via SAM3 and ensembles.

    Jupyter Notebook 2

  3. pedroreis2468/IIA pedroreis2468/IIA Public

    🏎️ AI Assistant for Formula Student Rules using RAG and Local LLMs.

    HTML 2

  4. pedroreis2468/ASM pedroreis2468/ASM Public

    🏛️ Distributed multi-agent system for smart university library management using SPADE/XMPP, FIPA-ACL protocols, and Contract Net negotiation, with real-time dashboard.

    Python 3

  5. pedroreis2468/AR2 pedroreis2468/AR2 Public

    🏎️ Autonomous Formula Student racing agent using Reinforcement Learning (SAC/PPO) in a 2D simulator with FS-AI rules, procedural tracks, and domain randomization.

    Jupyter Notebook 1

  6. GuilhermeLobo225/DAA GuilhermeLobo225/DAA Public

    🚦 Road Traffic Flow Prediction in Porto using Machine Learning.

    Jupyter Notebook 3