AI Master's student with an Engineering Physics background. I like building ML & GenAI systems end-to-end — from a clean data pipeline, to a model that trains, to an app someone can actually use.
- 🔭 Focus: deep learning · generative AI (RAG / LLMs) · multimodal & audio · data-driven engineering
- 🌱 Currently exploring: multi-agent systems, MLOps, and getting models from notebook to production
- 🎯 Looking for: a Summer 2026 internship in ML / AI engineering
- 📫 Reach me: guipinto2004@gmail.com
Academic, but real — competitions, full pipelines, and shipped prototypes. Grades on a 0–20 scale.
| Project | What makes it interesting | Stack | |
|---|---|---|---|
| 🥇 | AI-Generated Text Detection | 1st of 25 teams across all 3 submissions — telling human text apart from Anthropic / Google / Meta / OpenAI. From-scratch NumPy net → DistilBERT → an LLM stacking ensemble. · 19/20 | PyTorch DistilBERT LLMs |
| 🥉 | Porto Traffic Flow Prediction | 4th of 57 teams in a university Kaggle competition — a full CRISP-DM pipeline that held up between local CV and the private leaderboard. · 18.2/20 | scikit-learn Pandas |
| 🎧 | SmartHandover | Multimodal late-fusion (Whisper + RoBERTa + VADER + wav2vec 2.0 → XGBoost) that detects frustration in voice support and hands off to a human — 95.9% conversation-level recall. | Whisper RoBERTa wav2vec2 |
| 🤖 | FS Rules AI Assistant | A 100% offline RAG agent for Formula Student regulations: ask in plain language, get answers grounded in the real rulebook with citations. · 18/20 | RAG Ollama Gemma 3 12B |
| 🛢️ | The Price of War | An interactive D3.js data story tracing global oil flows — 8 official datasets stitched together by an auto-updating pipeline (map, Sankey, a calculator you can play with). | D3.js v7 GitHub Actions |
| 📡 | Smart Study Room | End-to-end IoT: two ESP32 nodes + YOLOv11 vision feed a Flask API, a web dashboard, and a native Android app, with live comfort forecasting. | ESP32 YOLOv11 Android |
➕ Also: a Kimball data warehouse with NLP sentiment, and a bi-criterion route optimizer (time vs. CO₂). More in my pinned repos.
Languages
ML & Deep Learning
GenAI & Data
Web · Tooling · IoT
