A.01
Software architecture
Boundaries, contracts, asynchronous flows, and evolutionary decisions.
- APIs
- MCP
- SSE
- Docker
SOFTWARE ENGINEERING // ARCHITECTURE // APPLIED AI
I build systems that turn complexity into clarity.
Software Engineer specializing in architecture, intelligent systems, and applied AI. Over nine years turning real problems into reliable software.
01 // MISSION PROFILE
Engineering driven by decisions, not trends.
My specialty lives at the intersection of architecture, product, and execution.
I have worked in software development for over nine years, including payment technology. I build with Python, PHP, and TypeScript, connecting APIs, data, web applications, and artificial intelligence with a focus on privacy, operations, and business outcomes.
- 9+
- years in engineering
- 03
- featured systems
- PT·EN·ES
- international presence
02 // CAPABILITY MATRIX
From architecture to production software.
Skills demonstrated in complete systems, with explicit technical decisions, constraints, and operations.
A.02
Intelligent systems
RAG, embeddings, vector search, and responsible model integration.
- LLMs
- RAG
- pgvector
- ML
A.03
Full-stack engineering
Interfaces, services, persistence, and automation as one coherent solution.
- Angular
- Node.js
- FastAPI
- PostgreSQL
03 // SELECTED SYSTEMS
Projects presented as engineering decisions.
Three case studies exploring knowledge, risk, and optimization from different angles.
SYS.01LIVE CASE STUDY
RAG · LOCAL FIRST · MCP
Knowledge Hub
A workspace to organize knowledge, retrieve context with confidence, and connect AI agents.
- Challenge
- Documents, notes, and references lose value when scattered and impossible to retrieve by meaning.
- Decision
- A RAG architecture with Angular, FastAPI, PostgreSQL/pgvector, local inference, and an access-controlled MCP server.
- Outcome
- A knowledge base searchable by semantics and grounded questions while preserving data control and privacy.
- Angular
- FastAPI
- PostgreSQL
- pgvector
- MCP
- Local LLM
SYS.02LIVE CASE STUDY
MACHINE LEARNING · XAI · RISK
Credit Card Fraud Detection
A transaction analysis lab combining sequence-based detection and AI-assisted interpretation.
- Challenge
- Prioritize anomalous behavior without turning an automated score into an opaque, final decision.
- Decision
- Combine temporal modeling, explainability, and generative agents to support — not replace — human analysis.
- Outcome
- An applied environment for investigating risk, interpreting signals, and experimenting with assisted decisions.
- Python
- FastAPI
- PyTorch
- scikit-learn
- PostgreSQL
- LangGraph
SYS.03LIVE CASE STUDY
GENETIC ALGORITHM · JOBS · SSE
Knight's Tour
A full-stack application for observing genetic algorithms explore the knight’s tour.
- Challenge
- Search for a 64-move route in a combinatorial space subject to stagnation and local minima.
- Decision
- Run populations as jobs, stream progress with SSE, and combine mutation strategies with Warnsdorff’s heuristic.
- Outcome
- A visual, measurable experience for tracking convergence, diversity, and stagnation recovery.
- TypeScript
- Node.js
- SSE
- SQLite
- Genetic Algorithms
04 // EXPERIENCE LINE
Knowledge built between software and real-world problems.
Software Engineering
gsurf · Payment technology
Software development in the payments domain, combining technical knowledge, reliability, and business problem-solving.
Software Engineering with Applied AI
UNIPDS · Postgraduate program
Advanced study in architecture, artificial intelligence, and practical intelligent systems.
Computer Science
University of Southern Santa Catarina
Degree completed with a taxonomic identification system based on dichotomous keys.
3rd place in programming contests
Academic competitions
Recognition for algorithmic reasoning and problem-solving under time constraints.
05 // ENGINEERING NOTES
Principles extracted from practice.
Short reads on decisions that cut across the featured systems.
NOTE.RAG
Reliable RAG begins before the answer
Ingestion quality, embedding identity, and source traceability matter as much as the generative model.
↗NOTE.RISKA score is not a decision
In risk scenarios, a model becomes more valuable when people can interpret its uncertainty and signals.
↗NOTE.SEARCHStagnation is a signal too
Evolutionary algorithms must observe diversity and change strategy when the population stops exploring.
↗06 // OPEN CHANNEL
Let’s talk about systems worth building.
Architecture, applied AI, software engineering, or a good technical question: the channel is open.