Luiz Fernando

Projects

A showcase of my technical projects and contributions across different domains.

Diet Pandas

Python Library

Automated memory optimization library for Pandas DataFrames. Intelligently reduces memory usage by 50-80% through aggressive type downcasting while preserving data integrity. Published on PyPI with comprehensive documentation.

πŸš€ Highlights

  • Published on PyPI - available via pip install diet-pandas
  • 50-80% memory reduction on real-world datasets (tested on 4.3M+ rows)
  • Drop-in replacement for pandas.read_csv with automatic optimization
  • Full documentation with MkDocs, benchmarks, and CI/CD pipeline
Python Pandas NumPy PyPI MkDocs
Rubber Duck Detection Project Demo

Rubber Duck Detection

Computer Vision

Intelligent rubber duck detection system powered by YOLO deep learning architecture. Specifically designed for identifying rubber ducks in various environments with high accuracy and real-time processing capabilities.

πŸš€ Highlights

  • YOLO-based object detection for precise localization
  • Robust to various lighting and weather conditions
Python YOLO OpenCV
View on GitHub
Real-Time Anomaly Engine Demo

Real-Time Anomaly Engine

Streaming & Observability

Production-ready real-time anomaly detection engine for e-commerce logs. It combines high-throughput Kafka ingestion, Spark structured streaming, and TimescaleDB for metrics storage, plus LLM-powered insights.

πŸš€ Highlights

  • End-to-end streaming pipeline with Kafka and Spark
  • RAG pipeline + embeddings for LLM-guided root cause insights
  • Automatic anomaly detection with alerting
Python Spark Kafka LangChain
View on GitHub
Soletras Project Demo

Soletras

Full-Stack NLP

Brazilian word searcher system that helps people complete the Brazilian SpellingBee. Features advanced filtering and search capabilities powered by SpaCy NLP.

πŸš€ Highlights

  • SpaCy NLP to filter conjugated verbs and plurals
  • 550,633+ Brazilian Portuguese words
  • Modern React 18 + Vite frontend
Python Flask React SpaCy
View on GitHub
PUMLE Project Demo

PUMLE

Data Engineering

Innovative data pipeline for geological carbon storage in Brazilian reservoirs. Transforms complex simulation outputs into streamlined multidimensional arrays.

πŸš€ Highlights

  • Multi-format export (NumPy, Zarr, MAT-files, CSV)
  • Optimized for large-scale geological simulations
Python NumPy Zarr
View on GitHub
Car Plate Detection Project Demo

Car Plate Detection

Computer Vision

Intelligent license plate recognition system powered by YOLO deep learning architecture. Designed for Brazilian vehicle registration patterns with high accuracy.

πŸš€ Highlights

  • YOLO-based object detection for precise plate localization
  • Character recognition optimized for Brazilian formats
Python YOLO OpenCV
View on GitHub
Michael Scott Generator Project Demo

Michael Scott Speak Generator

NLP

Text generation system powered by GPT-2 transformers architecture. Generates text in the style of Michael Scott from "The Office", leveraging advanced NLP techniques.

πŸš€ Highlights

  • Fine-tuning on character-specific dialogue
  • Contextual understanding of "The Office" episodes
Python Transformers PyTorch
View on Hugging Face
Motion Coach Project Demo

Motion Coach: Gym Assistant

Computer Vision

Innovative gym assistant application that leverages computer vision to provide real-time feedback on exercise form and technique. Achieved F1-score of 0.92.

πŸš€ Highlights

  • Real-time pose detection and tracking
  • Form analysis using keypoint detection
Python FastAPI MediaPipe
View on GitHub
Bin Packing Project Demo

Bin Packing Optimizer

Algorithms

High-performance solution to the Variable-Sized Bin Packing Problem (VSBPP). Built from scratch in C++ to achieve optimal space utilization with minimal computational overhead.

πŸš€ Highlights

  • Optimized C++ implementation for maximum speed
  • Variable-sized container support
C++ Algorithms Optimization
View on GitHub
Branch and Bound Project Demo

Branch and Bound

Optimization

Mathematical optimization powerhouse implementing the classic Branch and Bound algorithm. Designed to solve complex integer linear programming problems with guaranteed optimal solutions.

πŸš€ Highlights

  • Classic Branch and Bound implementation
  • Integer linear programming solver
Python MIP Math
View on GitHub
NumPy from Scratch Project Demo

NumPy from Scratch

Low-Level

Ground-up recreation of NumPy's core functionality using pure C programming. Demonstrates deep understanding of memory management, array operations, and performance optimization.

πŸš€ Highlights

  • Pure C implementation for maximum control
  • High-dimensional array manipulation
C Memory Management
View on GitHub