AI Engineerbuilding real-world ML systems

I design and deliver end-to-end AI applications, from computer vision and LLM systems to production-ready APIs and interactive web apps.

Live demos included across the portfolio. Not just code samples.

Sidnei Almeida

Projetos em destaque

End-to-end AI systems with live demos.

RAG AI document chatbot for PDF analysis and summarization

RAG AI Document Chatbot for PDF Analysis and Summarization

Ativo Recente Destaque

Ask questions directly to documents and extract insights from PDFs without manual reading, enabling faster research and decision-making.

Pipeline RAG
Suporte a PDFs
LangChain FAISS FastAPI OpenAI
AI Movie Recommendation System with Semantic Search

AI Movie Recommendation System with Semantic Search

Ativo Recente Destaque

Discover relevant movies using NLP-based similarity, enabling more accurate recommendations beyond traditional genre-based systems.

BERT Similarity
TMDb + YouTube
BERT TMDb API Semantic Search REST API
Predictive Maintenance System for Industrial Equipment

Predictive Maintenance System for Industrial Equipment

Ativo Recente Destaque

Predict equipment failures using time series sensor data and LSTM models, enabling proactive maintenance and more reliable industrial operations.

Model Accuracy (Test Set): 95.2%
Sensor Inputs: 7 industrial signals
LSTM TensorFlow FastAPI Plotly
Computer Vision System for Visual Defect Detection

Computer Vision System for Visual Defect Detection

Ativo Recente Destaque

Detect defects in industrial products using deep learning segmentation models, enabling automated and consistent quality inspection workflows.

Model architecture: U-Net segmentation
Dataset: MVTec AD (industrial defect detection)
U-Net PyTorch OpenCV Anomalib
Real-Time License Plate Detection System (ALPR)

Real-Time License Plate Detection System (ALPR)

Ativo Recente Destaque

Real-time license plate recognition with computer vision for monitoring, access control, and traffic analysis.

Precisão: 99.69%
Recall: 99.19%
mAP@50: 99.49%
Mercosul
YOLOv8 OpenCV FastAPI PyTorch
FluxForecast — LSTM real-time liquid flow rate prediction for offshore risers

FluxForecast · Liquid Flow Prediction in Real Time with LSTM

Active Recent Featured

Monitor and predict liquid flow rates in offshore risers using an LSTM model, simulating a live telemetry stream with P05/P95 threshold-based anomaly detection.

7 normalised pressure channels
Live stream with fixed P05/P95 thresholds
Dataset: 7,074 pre-scaled samples
LSTM FastAPI Render Time Series
Autonomous Research System · Multi-Agent LLM Orchestration

Autonomous Research System Multi-Agent LLM Orchestration

Active Recent Featured

Autonomous research system that orchestrates multiple AI agents to gather, process, and synthesize information from scientific sources.

Multi-agent orchestration
Scientific sources + retrieval
LangChain LangGraph Groq RAG
Deep Reinforcement Learning Trading System

Deep Reinforcement Learning Trading System

Active Recent Featured

Quantitative trading system powered by deep reinforcement learning models to simulate, optimize, and execute strategy decisions with risk control.

Live balance & P&L
Open positions
Risk guard-rails
FastAPI · Render
Deep RL Trading FastAPI JavaScript

Habilidades

Sistemas de IA, LLMs e stack de implantação

Sobre Mim

AI Engineer focused on building and deploying real-world machine learning systems.

I design end-to-end AI applications across computer vision and LLM-based systems, including APIs, real-time inference pipelines and interactive frontends.

My work includes systems like industrial anomaly detection, document-based AI assistants (RAG) and real-time computer vision applications, all available through live demos.

I focus on practical impact, building AI systems that are reliable, scalable and ready for real-world use.

LinkedIn
Sidnei Almeida

Vamos Conversar

Open to AI Engineering roles, freelance projects, and collaborations. If you're building something with machine learning, computer vision, or LLMs, I'd be glad to connect and contribute.