Guilherme Ciccarelli

Multi-agent systems
& LLMs applied
to real problems.

CTO at bee6 . co-founder of Una Edu.

I build production AI systems " not demos. Credit risk, legal document intelligence, real estate pricing, adaptive education. Five years shipping into messy data, no clean APIs, clients who need results.

See my work WhatsApp Email
GitHub LinkedIn
Python Multi-agent LLMs RAG / FAISS FastAPI Google Cloud Run SQLite . BigQuery scikit-learn . XGBoost
5+
Years shipping AI systems
6
Industries served
3
Companies co-founded
Clients in Brazil & the US

What I've built

Production systems across AI, legal, real estate, finance, education, and restaurants. Messy data, real clients, real results.

Una Edu

Multi-agent platform generating personalized learning materials for autistic students. Teacher sets the subject; system outputs a full adapted lesson with text, images, and assessments. LLMs · Cloud Run · FastAPI.

unaedu.com.br

bee6

Brazilian AI consultancy serving clients in Brazil and the US. RAG systems, automation pipelines, analytics stacks, and e-commerce for 10+ clients across legal, hospitality, real estate, and finance.

bee6.com.br

Berimbau NYC

Analytics infrastructure for the best Brazilian restaurant in NYC. POS, payroll, and reservations unified into a single BI layer with LLM-powered reporting. Bee6's anchor US client.

Client project

Legal Document Intelligence

Document intelligence pipeline for a law firm. Audio transcription, vector retrieval, named entity extraction, and structured report generation. RAG · NLP · local inference.

Full case study

Urban Space

Real estate intelligence platform for the São Paulo market. ML pricing model, geospatial enrichment, and lead mining from public records. Built for real estate brokers.

Full case study

Solo Inteligente

Land parcel intelligence for rural property research in Brazil. RAG pipeline over agricultural and registry documents to automate rural due diligence.

Full case study

Cicor

Credit risk automation for a factoring operation. ML scoring, document parsing, and daily data ingestion into a cloud analytics layer. Runs fully unattended.

Full case study

bquant

Quantitative research toolkit for Brazilian macro and real estate data. Time series modeling, volatility analysis, and data pipelines from public sources. Personal research project.

Full case study

GuiOS

Personal AI operating system. Multi-agent system over a local database tracking finances, health, travel, and projects. Accessible via Telegram. Built to reduce friction, not add it.

Full case study

What I do well

Building AI systems and data pipelines that run in production, not notebooks.

🤖

LLM Systems & RAG

Multi-agent architectures, retrieval-augmented generation, vector search with FAISS, prompt engineering, and local inference on constrained hardware. From prototype to Cloud Run deployment.

⚙

Automation & Pipelines

Cron-driven data pipelines, document processing workflows, Whisper transcription, web scraping at scale, and multi-step agentic systems that run without human intervention.

📉

Time Series & Econometrics

ADF/KPSS stationarity tests, ARIMA/ARCH modeling, ACF/PACF analysis, rolling-window volatility, macroeconomic indicators from BCB/IBGE/IPEA. Applied to real estate, financial markets, and credit risk.

🌎

Geospatial Analysis

Urban data pipelines with GeoSampa, INCRA, CAR, and SIGEF. Spatial joins via cKDTree, QGIS for zoning overlays, and enrichment of 11M+ IPTU parcels with socioeconomic, HDI, and transit proximity data.

🧮

Machine Learning

Supervised and unsupervised models for credit scoring, property pricing, and land classification. Random Forest, XGBoost, PCA + KNN. scikit-learn pipelines from raw data to production inference.

📊

Data Engineering

SQLite and BigQuery architectures, pandas pipelines, BCB/CVM/B3 data ingestion, FipeZAP real estate indices, and custom analytics stacks that feed both internal tools and client-facing dashboards.

Education & Certifications

Formal academic background combined with applied technical certifications.

🎓

B.A. Social Sciences

Universidade de São Paulo (USP)
2009 . 2012
🌎

Computational Social Science Methods

University of California, Davis
Jul 2023 . Credential: BMJGNQS8CSPR
🤖

Open Source LLMs & Secure AI with RAG

Udemy
Oct 2025
🧠

Introduction to Machine Learning

Universidade de São Paulo (USP)
Mar 2022
🐍

Python for Data Analysis

Universidade de São Paulo (USP)
Mar 2022
🐳

Credit Risk Modeling in Python

Udemy
Oct 2021
📈

SQL for Data Science with Google BigQuery

Udemy
Jan 2021
📊

Data Analysis with Pandas and Python

Udemy
Sep 2020

AI Engineer.
Builder. Founder.

I build AI systems that work on real problems " not demos.

Over the past 5 years I've shipped production systems across credit risk, real estate pricing, legal document intelligence, restaurant analytics, and adaptive education. Most of them involved messy data, no clean APIs, and clients who needed results, not notebooks.

Currently CTO of bee6, a Brazilian AI consultancy with clients in Brazil and the US, and lead developer of Una Edu " an AI platform that generates personalized learning materials for autistic students, built because I'm the father of an autistic child.

My current focus is multi-agent architectures and LLM pipelines applied to real-world data problems. Open to conversations about AI engineering, applied research, or anything at the intersection of data and real-world impact.