Humberto Lomeu, founder of Retrieverworks, AI and data engineer

Humberto
Humbertojust now

These messages are generated by an AI agent. Not me typing in real time.

About

Nine years of advisory and production AI.

I started in AI at EY in 2017, then spent four years at a Brazilian AI startup, later acquired by one of the country's largest banks, leading delivery for health insurers, banks and retailers. Most recently I built deep learning forecasting models for TotalEnergies while completing an MSc in Computer Science.

Today I run Retrieverworks, the AI and automation studio I founded in York: AI agents, workflow automation and bespoke builds, taken from prototype to production.

Client work runs through Retrieverworks. Freelance and contract work through Upwork.

Humberto Lomeu, portrait
Studio

Big-firm AI, sized for a small business.

Retrieverworks is the AI and automation studio I founded, based in York and working with businesses across the UK. I build AI agents, workflow automation, systems integration, data migration and bespoke web applications, scoped honestly and shipped to production.

Retrieverworks
Career

A small, deliberate track record.

  1. 2026 — now
    Founder, AI & Data Engineer
    Retrieverworks

    Founded Retrieverworks, an AI and automation studio based in York. Through the studio I build AI agents, workflow automation, and bespoke web applications for small businesses.

  2. 2024 — 2026
    Deep Learning Researcher
    COPPETEC · TotalEnergies

    Built deep learning models for offshore wind speed forecasting, combining regional weather model outputs with SAR satellite data. The work became my MSc dissertation at COPPE/UFRJ, defended February 2026.

  3. 2020 — 2024
    Data Science and AI Manager
    Kunumi

    Led AI delivery at a Brazilian consultancy across health insurance, banking, retail, and infrastructure. Managed up to 18 data scientists, engineers, and coordinators while staying hands-on technically; the portfolio I owned contributed up to 40% of company revenue at peak.

  4. 2017 — 2020
    Consultant, Data & Analytics
    EY (Ernst & Young)

    Financial services clients (banks, insurance, pension funds): data analytics, automation, early machine learning. Self-taught Python in 2018 when no one else on the team was using it, then built a web scraping and NLP pipeline that classified 100,000+ court documents and led a workstream on LGPD data privacy compliance.

Hobby projects

Things I build for myself, outside client work. They keep my deployment, CI and observability habits sharp.

TfL Monitor

Live

Live Transport for London feeds landed in a dbt-modelled Postgres warehouse, with a LangGraph agent on AWS Bedrock that answers questions from SQL over the warehouse or RAG over TfL's strategy documents.

dbt·Airflow·LangGraph·AWS Bedrock·FastAPI

Tech stack