São Paulo, Brazil

I build AI for businesses that run on spreadsheets and WhatsApp.

Co-founder and CTO at bee6, co-founder of Una Edu. Social scientist by training, self-taught engineer since 2020. Before that I spent six years analyzing credit risk by hand, which is how I got into automating things.

Portrait of Guilherme Ciccarelli
Guilherme Ciccarelli, 2026
Now

Running engineering at bee6: AI, data pipelines and internal tools for clients in Brazil and the US.

Also

Building Una Edu, adapted school materials for autistic kids. I'm the father of one.

Stack

Python, LLM agents, RAG, FastAPI, Cloud Run, BigQuery, SQLite, scikit-learn, faster-whisper.

Selected work

Systems I designed and wrote, from client work to personal research.

  1. 01

    Una Edu

    A teacher picks the subject and the system writes the lesson, exercises and test, adapted to each autistic student's profile and built around their hyperfocus. Ready to print.

    Education
    Live
    ↗
  2. 02

    Legal RAG

    Contracts, court records and hearing audio go in; cited legal briefs come out. RAG for a law firm, with the whole document base indexed on the lawyer's own machine.

    Legal
    Client
    →
  3. 03

    Berimbau Brazilian Table

    The data side of two Brazilian restaurants in New York. Sales, payroll and reservations land in one BI dashboard every day, plus the website and a seat-booking app for their live music floor.

    Hospitality, NYC
    Live
    ↗
  4. 04

    Cicor

    Credit scoring, document parsing and daily data ingestion for a factoring company, running unattended. Also Duplinet, the portal where their clients sell receivables.

    Credit
    Live
    →
  5. 05

    Urban Space

    A pricing model and lead mining for São Paulo real estate, on top of 11M+ IPTU parcels enriched with income, HDI and transit data.

    Real estate
    Research
    →
  6. 06

    Solo Inteligente

    Finds São Paulo lots that the 2023 Master Plan lets you build denser on, before the market reprices them. The job scouts used to do lot by lot on Google Maps.

    Land
    Research
    →
  7. 07

    bquant

    My research toolkit for Brazilian macro and real estate series: stationarity tests, ARIMA/ARCH, rolling volatility, public data pipelines from BCB, IBGE and IPEA.

    Quant
    Personal
    →
  8. 08

    GuiOS

    A multi-agent operating system for my own life: finances, health, projects, friends. It stored everything and I never asked it anything, so I shut it down. The write-up says why.

    Personal
    Retired 2026
    →

About

I spent six years analyzing credit risk at a factoring company, mostly with spreadsheets and gut feeling. I got bored and suspected the whole job could be automated. In 2019 I co-founded Duplify with two friends to rebuild the factoring process from scratch, and in 2020 I wrote my first line of Python to build its credit model.

I learned the statistics and machine learning on the way. In 2022 a post-doctoral researcher at Unicamp saw that credit project and offered me a Master's position. I didn't take it, but it told me I was on the right track.

In 2023 I co-founded bee6 with four partners. We build AI and data systems for a law firm, restaurants in New York, a factoring operation and schools. Most of these clients run on spreadsheets and WhatsApp, so a good part of the job is getting reliable data out of a messy operation before any model gets near it.

My degree is in Social Sciences, from USP. It turned out to be useful: most data problems are people problems with a CSV attached. I'm also the father of an autistic boy, which is why Una Edu exists.

Education
B.A. Social Sciences, Universidade de São Paulo, 2009–2012
Computational Social Science Methods, UC Davis, 2023
Companies
Duplify (2019), bee6 (2023), Una Edu
Contact

Email or WhatsApp.
WhatsApp is faster.

guicicca90@gmail.com