fullstack → ai engineer
Marcus Rodrigues
I build generative AI applications, from RAG to agents and LLM integration, connected to real systems in Java, .NET and Angular. I came from law, which taught me to understand the problem before writing the solution.
About
I started in law. I spent three years at the Rio de Janeiro Bar Association (OAB/RJ), where I taught lawyers how to use the courts' electronic filing systems and saw up close how complex legal processes turn into poorly built software. At some point I decided I'd rather build the tools than put up with them.
Today I'm a full-stack developer at Vibetex, responsible for the AI layer of the products. On the project for CEDAE, Rio's state water utility, I joined a platform already in production with over 10,000 field inspections and added automatic classification, summaries, inconsistency detection and a natural-language chat to query the data. The project closed with a technical capability certificate from the client. I'm now leading the build of a support channel with LLMs, RAG and semantic search, on top of Java and .NET APIs and Angular interfaces.
On the side I'm building Havrus, my own AI product for the legal market, where my two backgrounds meet. To me, AI only creates value when it's treated as real software: well-designed APIs, reliable data, clean code and answers you can evaluate and audit. I'm also completing a degree in Systems Analysis and Development, due in 2027.
Work
Four projects that show the path: AI in production for a public client, my own product, a hackathon win, and this site.
AI for field inspections (CEDAE)
AI layer on a production platform with over 10,000 inspections: automatic classification, summaries, inconsistency detection and a natural-language chat. Closed with a technical capability certificate.
LLMs · .NET · Java · Angular · SQL Server
Full case study coming soon
Havrus
RAG-based AI engine for law firms: case-law analysis, identification of arguments and risks, and drafting support. Multi-tenant, privacy by design. In progress.
Java · Spring Boot · Angular · LLMs · RAG
Read the case study
Coti × Criare Hackathon 2025
1st place in 24 hours: a quoting and negotiation platform for companies and suppliers, with live negotiation over WebSocket and JWT with rotating refresh tokens.
Java · Spring Boot · Angular · PostgreSQL · Docker
Read the case study
This site
Portfolio built spec-first, with its own identity and two languages written separately.
Next.js · TypeScript · Tailwind · Framer Motion
Read the case study
Stack
- AI
- LLMs · RAG · Embeddings · Vector search · Agents · Prompt engineering
- Backend
- Java · Spring Boot · C# · .NET · REST APIs
- Frontend
- Angular · TypeScript · Next.js
- Data and infra
- SQL Server · PostgreSQL · MongoDB · Docker · Git
Contact
Open to conversations about AI engineering and products in any field. If the problem happens to be legal, even better: it's ground I know from both sides.