AI student at UNIFEI (Universidade Federal de Itajubá) with hands-on experience in machine learning research and Python backend development.
Author of dNATY — an open-source neural network compression engine that achieves −46.5% FLOPs and 1.6× inference speedup with no retraining, commercialized as a SaaS product.
class PedroVergueiro:
def __init__(self):
self.education = "B.Sc. Artificial Intelligence @ UNIFEI (2024–2028)"
self.company = "Vergueiro Tech (MEI) — Founder & Developer"
self.research = "Neural Architecture Search + Continual Learning"
self.open_to = ["AI/ML Internship", "Python Backend", "Research"]
def flagship_project(self):
return {
"name" : "dNATY",
"flops" : "-46.5%",
"speedup" : "1.6x",
"cl" : "6.9x more efficient than EWC",
"url" : "github.com/pedrovergueiro/dNaty",
}Compress any PyTorch model with one function call — no retraining required.
from dnaty import compress
result = compress(model, train_data, target_flops=0.5)
# → -46.5% FLOPs | 1.6× faster | accuracy preserved| Metric | Result |
|---|---|
| FLOPs reduction | −46.5% |
| Inference speedup | 1.6× |
| vs EWC (continual learning) | 6.9× more efficient |
| Retraining required | None |
Stack: Python · PyTorch · NumPy · FastAPI · SQLAlchemy · Celery · Redis · Docker · GitHub Actions · pytest
dNATY — Neural Compression Engine
Open-source · SaaS · −46.5% FLOPs · 1.6× speedup · No retraining
Código Brasil — Python Education
Teaching Python to 13 students · Democratizing programming education in Brazil
REST API for NLP-based text sentiment classification · Python · FastAPI
Collaborative filtering recommender · Python · Scikit-learn · TF-IDF
Full CRUD REST API · Python · FastAPI · SQLite
Continuous Improvement Intern — ACG Capsules (Oct 2025 – Mar 2026)
Developed production management & factory operations system for a pharmaceutical multinational.
Founder & Developer — Vergueiro Tech (MEI) (2024 – Present)
4 production websites delivered · dNATY SaaS platform built and deployed.
