Research & Applied Domain AI
Research & Legal AI
2023–2024 (research) · 2025–2026 (applied) · Fortaleza · Brazil → FI Group (Brazil / Spain footprint)
A two-phase arc — legal-document research at Unifor / NCDIA, then applied retrieval, precedent, and prediction systems at FI Group. The chapter that taught João how to work in structured, high-context, high-precision domains.
- Legal AI
- Classification
- Retrieval
- Domain intelligence
- Research foundation
Period
2023–2024 (research) · 2025–2026 (applied)
Significance
A foundational thread in João's trajectory: research discipline, structured reasoning, and domain-specific AI system design, carried from academic work into applied internal-AI infrastructure.
Why it matters
Shows João's ability to move from academic investigation to applied AI systems in high-context domains where grounding, structure, and precision matter far more than generic model outputs.
Story arc
How this chapter unfolded.
- 01
Research foundation
Learning to structure complex legal information.
João's early legal-AI work at Unifor / NCDIA focused on classification, hybrid retrieval, and the discipline of turning messy legal material into structured, machine-readable problems. The rigor came first; the systems came later.
- 02
Method
Precision mattered more than generic generation.
The work trained the instinct that later became central to João's AI systems: outputs need structure, traceability, and evaluation — especially in high-context domains like legal, policy, and business workflows. PromptBreeder + DSPy beat expert-crafted prompts and LegalBERT-pt; the lesson was that careful evaluation wins over confident-sounding output.
- 03
Continuation
From classification to grounded intelligence.
More than a year later, João's FI Group work extended that foundation into retrieval, precedent analysis, prediction workflows, and agentic systems for domain-specific decision support. Same theme; different chapter; different employer. The thread is continuity, not timeline collapse.
Curated media
A few moments that tell the story.
The research environment at Unifor — where the legal-AI work began with classification, hybrid retrieval, and prompt-optimization research backed by FUNCAP.
Academic context tied to the NCDIA group — the discipline of high-context, high-precision AI work that became this chapter's foundation.
Phase two — the FI Group Lei do Bem Precedent Analysis Agent (Oct/Nov 2025 – present). Agent workflow, RAG microservice, vector store, LLM, source-cited answers. The applied output of the research instinct on a much larger surface.
Phase two — the FI Group Technical Intelligence Prediction Agent. Extraction, embeddings, ML artifacts, structured advisory output. Decision-support intelligence, not loose chatbot answers.
Proof and links
External signal.
article
Unifor interview on João's international technology path
A later Unifor interview frames João's Stanford experience, applied AI work, Huawei recognition, and FUNCAP legal AI research as part of a broader international trajectory in technology.
social
LinkedIn profile
Public professional profile describing João as an AI engineer focused on scalable RAG, agentic AI, and computer vision, with Stanford and EuroTech listed among his signals.
Continue the arc