Matheus de Camargo Marques
Specialist in concurrent and distributed systems with Elixir. I transform complex architectures into scalable and performant solutions.
Specialties
Concurrent & Distributed Backend
Developing highly scalable systems using Elixir/OTP. Experience with microservices, message queues, and resilient architectures.
Database Optimization
Performance tuning, strategic indexing, query optimization, and data modeling for high-demand applications.
Legacy Maintenance & Improvement
Refactoring legacy code, modernizing architectures, and implementing best practices without interrupting operations.
Academic background
Graduate degree (Stricto Sensu)
In progressUTFPR - Curitiba
External Graduate Student (Stricto Sensu)
Computer Engineering
Bachelor
Bachelor's degree in Computer Science
Degree
IT technician
Technical training
Logistics technician
Technical training
Technology Stack
Backend
Frontend
Mobile & Others
Projects & Research
NeuroStack-RAG
A Bio-Inspired Cognitive Architecture for LLMs
Executive Summary
Traditional Retrieval-Augmented Generation (RAG) systems face a critical problem: entropy. Over time, the accumulation of unstructured data generates noise, slowness, and increases the rate of hallucinations. NeuroStack-RAG is an architectural response to this problem, simulating the biological memory mechanisms of the human brain—active forgetting, memory consolidation, and associative learning—to create a knowledge base that not only grows but evolves and becomes more intelligent with use.
The Problem
In conventional data architectures, knowledge is static. A document inserted today has the same "weight" in a vector search a year from now, regardless of its actual utility. This creates a linear growth of noise, where relevant information is suffocated by obsolete data, degrading the language model's performance.
The Solution: Data Biomimicry
NeuroStack abandons the idea of a "static database" in favor of a living memory system. It introduces the concept of an information life cycle, where each piece of data must "fight" for its survival based on utility and reinforcement. The architecture operates on three conceptual pillars:
1. Three-Layer Memory Hierarchy
Like the brain, the system segments data by "age" and semantic complexity: Eden Space (Working Memory), Survivor Space (Episodic Memory), and Tenured Space (Semantic Memory).
2. Augmented Synapse (Hebbian Learning)
"Cells that fire together, wire together." The system applies this principle to data, weighting relevance by access history. Useful memories become "synaptically strong," while irrelevant ones weaken.
3. "Sleep" Cycles and Offline Consolidation
Asynchronous processes analogous to human sleep for maintenance. The Efficiency Phase (NREM) optimizes searches, and the Synthesis Phase (REM) "dreams" about the data to find connections and generate new insights for the Knowledge Graph.
Technical Differentiators
- Active Forgetting (Semantic Garbage Collection): Drastic reduction in storage costs and noise by automatically eliminating obsolete data.
- Contextual Retrieval: The system prioritizes what is important, not just what is similar.
- Anti-Hallucination: By consolidating facts into knowledge graphs and validating contradictions during offline processing, the system increases the reliability of its responses.
This project is an ongoing exploration of how we can make AI more efficient and closer to human cognition. Interested in discussing cognitive systems architecture? Get in touch.
Current Study Areas
Recommendations
Marcos Vinícius
Software Engineer | Java | Kotlin | Spring Boot | Docker
I had/have the privilege of working with Matheus and I can say that he is a rare professional. Besides his great technical capacity, which is always impressive, Matheus stands out for his ability to lead teams in an inspiring way. He is a true facilitator, always ready to listen, support, and find solutions, always with great humility. His accessible and collaborative posture creates a motivating and productive work environment.
Daniel Oliveira
Golang Developer
Matheus is an exceptional professional, who stands out for his proactivity, strategic vision, and constant search for improvements. He is always attentive to new technologies and market trends, but without neglecting the positive aspects of consolidated solutions. Furthermore, he is a great colleague, who shares his knowledge and contributes to the team's growth.
Emanuel Kidoguchi
Fullstack Developer | Elixir | Phoenix | LiveView
Matheus has solid knowledge in Erlang/Elixir, besides being passionate about the language. Highly discerning and focused on developing scalable, fault-tolerant software, he is an excellent professional, especially in the context of OTP with Elixir and Erlang.
Mikael Santos
Key Account Manager
Matheus is excellent at organization, a very focused professional. As a programmer, he has an incredible mind for absorbing/replicating knowledge, due to his mastery of languages and experience.
Cleverson Matias
Fullstack Developer | Node.js | PHP | React | SQL
A great colleague, dedication is everything to Matheus, always focused on doing his best and with an impressive richness of detail. I really enjoyed working with him, great person! excellent professional!
Mateus de Matos Leal
Software Engineer
Persistence is the word that comes to mind when I think of Matheus. Above all, I was impressed with Matheus's ability to work hard on every project. And, of course, his gentle way of treating friends. Matheus would be a true asset to any position requiring programming and comes with my sincere recommendation.
I believe that each of us possesses unique gifts — and that we have the responsibility to put them at the service of the world. In my case, I found this calling in the world of distributed systems, with deep knowledge of the Erlang/OTP stack and the Elixir language.
I work with purpose. I enjoy solving complex problems — those that seem bigger than us — because I know I'm not walking this path alone. I teach what I learn, share what I build, and contribute to the community whenever I can — whether by mentoring, improving open-source code, or organizing meetups to foster connections.
I see technology as a powerful instrument — one that, in the right hands, can transform lives. That's why I strive for excellence. Not out of pride, but because I believe that giving our best is a form of gratitude. Soulless projects don't motivate me. I am guided by purpose, not just by code.
I have failed. I have hesitated. And it was in those moments that I grew the most. Today, I understand that challenges don't come to stop me — they come to shape me. Every obstacle is an opportunity to serve better, with more humility, faith, and courage.
Matthew 25:14-30 — the Parable of the Talents — reminds me daily of what is at stake: not to waste what has been entrusted to me.
If you feel called to go further, if you seek partners with purpose, and if you believe that technology can be a bridge — and not an end — it will be a privilege to connect with you.