Education
B.Sc. in Computer Science
University of São Paulo (IME-USP)
2020–2024
AI Engineer in São Paulo
Luiz Carlos
LLM systems, backend and automation.
I build production LLM systems, backend services, and the automation around them. I lead AI and automation at BWG; before that, I built agent pipelines at Auria AI that processed 3,000+ security incidents a day.
20+production AI agents built
3,000+security incidents processed daily
5insurers’ quote flows in one platform
01About
Engineering at the edge of data and AI.
I’m most useful where models, backend systems, and real business processes meet.
At BWG, I own AI and automation for a corporate health-insurance brokerage, from process discovery to deployment and adoption, reporting directly to the CEO. Each step of a workflow gets the right tool: an API integration, deterministic automation, LLM assistance, or human review.
Before that, at Auria AI, I built 20+ production AI agents in an incident-response pipeline handling 3,000+ B2B security incidents a day, along with the PostgreSQL data layer, BI dashboards, and on-call alerting around them.
Earlier, I migrated a legacy Ruby on Rails platform to Kotlin at IME-USP. I hold a B.Sc. in Computer Science from the University of São Paulo.
Languages
PortugueseNative
EnglishAdvanced
MandarinLearning
02Experience
Building, migrating, and leading.
Applied AI and production engineering, built on a foundation in research and technical leadership.
AI Engineer · Applied AI & Automation
BWG · HR Technology & Corporate Benefits · São Paulo, Brazil
- Own AI and automation initiatives for the corporate health-insurance brokerage, from process discovery to deployment and adoption, reporting directly to the CEO.
- Mapped the quoting and client-onboarding workflow step by step, classifying each step as API integration, deterministic automation, LLM assistance, or human review, and defined baseline metrics to evaluate the MVP.
- Built an internal platform for hospital-network coverage analysis, pricing automation on four insurers’ portals, quote-request forms for five insurers, and onboarding document generation. After the first demo, the team adopted it as their official quote-request channel.
- Built a Gemini extraction pipeline for client PDFs, spreadsheets, and scanned documents: every extracted field links back to its source, missing data is flagged instead of guessed, and pricing calculations stay outside the LLM.
- Deployed and operate the platform on Kubernetes, with CI/CD pipelines and automated backups.
TypeScriptReactNode.jsPostgreSQLGeminiKubernetesCI/CD
Backend Software Engineer
Auria AI · São Paulo, Brazil
- Built and maintained 20+ Python AI agents in the production incident-response pipeline, handling summarization, insight extraction, and automated decision-making for 3,000+ B2B security incidents per day.
- Bootstrapped and maintained the PostgreSQL/SQLAlchemy data layer across 36 production tables, shipping backend features end to end across multiple FastAPI services.
- Built Auria’s Metabase BI from scratch, including operational dashboards, KPI analyses, and threshold-based alerts used daily by incident-response leadership.
- Designed and shipped an on-call alerting bot that calls the responsible engineers through Twilio with AI-generated incident summaries, reducing overnight outage detection from 8+ hours to near-immediate.
PythonFastAPIPydantic-AIPostgreSQLSQLAlchemyMetabaseTwilio
Software Developer · Research Scholarship
Hub USP Inovação · IME-USP
- Maintained the innovation platform of AUSPIN (USP’s Innovation Agency), which catalogs 1,000+ patents, 3,000+ companies, 3,000+ professors, and 100+ innovation initiatives.
- Migrated the platform’s backend from a legacy Ruby on Rails system to Kotlin, redesigning domain models, REST APIs, automated tests, and Docker deployment infrastructure.
KotlinRuby on RailsREST APIsDockerAutomated testing
President & Member
USPCodeLab · IME-USP
- Served as President from January 2022 to August 2023, leading 20+ members; organized two university hackathons, secured sponsorships, mentored participants, and managed event budgets.
LeadershipMentorshipEvent operationsWeb development
03Skills
A production-minded toolkit.
Backend-first, with the AI, data, and frontend range needed to take a product from an idea to production.
Backend
Core engineeringPythonFastAPIasyncioPydanticSQLAlchemyNode.jsExpressREST APIsMicroservicesPytest
AI & LLMs
Applied intelligenceAI AgentsRAGEmbeddingsPrompt EngineeringStructured OutputsLangChainPydantic-AIGeminiOpenAIClaude
Data & Tools
Production infrastructurePostgreSQLSQLiteRedisDockerKubernetesGitMetabaseTwilio APIVector SearchCI/CD
Frontend
Product interfacesTypeScriptJavaScriptVue.jsReactHTMLCSS
04Projects
Ideas made useful.
Selected work across meeting intelligence and retrieval-augmented generation.
01Open sourceSince 2026
Memoir — AI meeting knowledge workspace.
An open-source full-stack app that turns meeting transcripts into searchable summaries, decisions, action items, participant profiles, and structured transcript segments. It detects Zoom and Google Meet formats automatically, with a Gemini fallback for messy or unsupported ones, and Pytest coverage protects parsing and core workflows.
FastAPIPydantic-AISQLAlchemySQLiteGeminiJavaScriptHTMLCSSPytest
View on GitHub
02Capstone2024
Divination — RAG assistant for Dungeon Masters.
A RAG assistant grounded in the official 2024 D&D Free Rules using LangChain, vector search, and GPT-4o. Built with a FastAPI backend and Vue 3 frontend as my B.Sc. capstone project at IME-USP.
PythonFastAPIVue 3LangChainGPT-4oVector Search
View project
05Contact
Let’s build something solid.
Happy to talk about LLM systems, process automation, or something you’re building. Based in São Paulo, reachable from anywhere.