ricini.dev
ENPT

Luiz Ricini

Full-stack engineer shipping AI products end to end.

I build the agent layer, the product around it and the infrastructure underneath, using React and Next.js, Node.js, Python and PostgreSQL. Now at Decode Data; before that at Rendair, the AI rendering platform architects use worldwide.

  • HEAD → main
  • open-to-work

Author: Luiz Ricini <luizricini1@gmail.com>Palotina, PR, Brazil · working remote

Email meLinkedInGitHubDownload CV
$ git log --merges --oneline
  1. 2026-09regulasRegulas: AI tax assistant, waitlist open
  2. 2026-09rcgiCarbon registry with traceability graph
  3. 2026-08agroidRural-credit compliance engine
  4. 2026-07agentsSecure file attachments in agent chat
  5. 2026-06agentsPer-client access control for AI agents
  6. 2026-06firewatchFarm onboarding from .kmz files
  7. 2026-05rendairAssets Library with bulk actions
  8. 2026-02rendairGuided onboarding to the first AI render

Selected work

Products I built and shipped to production, and what I owned in each.

Rendair

Rendair · United States

SaaS where architects around the world turn sketches and 3D models into photorealistic renders with AI. I shipped features across the React app and the serverless AWS API.

  • Built the Assets Library end to end: search and filters, folders with drag and drop, and multi-select bulk actions.
  • Replaced full-size previews with a custom thumbnail pipeline, cutting preview rendering cost by around 60%.
  • Shipped the guided onboarding that takes new users to their first AI render, and the backend for the Canvas workflow.
  • Added admin moderation to the public community feed and delivered production fixes across web and API.

AI data agents

Decode Data · Brazil

A multi-client platform of AI data agents: people ask questions in plain language and get answers, charts and dashboards from their own data. Python (FastAPI + Agno) behind a Next.js product.

  • Built the agent’s chat and interactive canvas: session management, RAG over pgvector, and guardrails for PII and prompt injection.
  • Designed per-client role-based access control, so an agent only ever reaches the data its user is allowed to see.
  • Shipped secure file attachments in the chat and a canvas editor with alignment guides and undo/redo.
  • Built a data anonymization pipeline that protects sensitive information while keeping datasets usable for analysis.

Alerta de Incêndio & AgroID

Decode Data · client work

Wildfire monitoring and rural-credit compliance for Brazilian agribusiness: farms, fire hotspots, satellite rasters and the regulatory checks banks and insurers must run.

  • Farm onboarding from .kmz files, with the property boundaries drawn on an interactive map.
  • Compliance checks for Brazil’s rural-credit and crop-insurance rules (CMN 5.267/2025, CNSP 485/2025), with an audit trail for every result.
  • Risk analysis fed by public satellite data on fire risk and land use.

RCGI-USP Carbon

Decode Data · for RCGI, University of São Paulo

Greenhouse-gas accounting and a carbon registry: companies and cities compute emissions under IPCC 2006, GHG Protocol and GPC 1.1, submit them for validation, and register carbon projects and credits.

  • Rebuilt a no-code prototype as an owned product on Next.js, tRPC, Drizzle and PostgreSQL.
  • Emissions calculator following IPCC methodology, with every figure traceable to its emission factor.
  • Municipal inventory dashboard, mitigation scenarios, and a public carbon registry that traces each credit back to its project.

Regulas

My own product · launching soon

An AI assistant for Brazilian accountants working through the 2026 tax reform, grounded in official sources and treating legal validity dates as a first-class concept. The waitlist is open.

  • Ingestion pipeline that turns laws and decrees into versioned provisions with validity rules and pgvector embeddings.
  • Hybrid retrieval (vector + full-text, merged with reciprocal rank fusion) feeding a Claude agent that cites the exact provision behind each answer.
  • FastAPI backend with chat sessions and usage quotas; Next.js chat with streamed responses and Google sign-in.

Havana & HerAI

Epicure · France

Two AI-powered iOS apps shipped end to end for Epicure: backend, custom AI APIs, the Swift front end and App Store releases.

  • Havana: a team of personal AI agents, such as a travel planner, a career coach and an astrologer, each with its own persona and prompts.
  • HerAI: an assistant dedicated to women, covering well-being, planning, writing and life coaching.
  • Custom AI APIs and real-time conversation flows built for fast responses, with secure third-party integrations.

Experience

Read it like a log: newest on top.

  1. merged

    Rendair

    Full-stack Software EngineerFull-time · RemoteCalifornia, United States

    • Thumbnail pipeline that cut preview rendering cost by around 60%; Assets Library, onboarding and Canvas workflow.
    • React and Node.js on AWS (Lambda, SQS, S3) for a platform architects use worldwide.

  2. ongoing

    Decode Data

    Full-stack Software EngineerPart-time · RemoteUberlândia, Brazil

    • AI agent layer with Agno: chat and canvas, session management, RAG over pgvector, PII and prompt-injection guardrails.
    • Product work across client platforms in agribusiness compliance, wildfire monitoring and carbon accounting.

  3. merged

    Epicure

    Mobile Software EngineerFreelance · RemoteParis, France

    • Launched Havana and HerAI on the App Store, from backend and AI APIs to the Swift and React Native front end.

  4. merged

    Onedev

    Software DeveloperFull-time · HybridPalotina, Brazil

    • Built a messaging platform with WhatsApp bot automation on Next.js, NestJS and PostgreSQL.
    • Cut execution time of core queries by 80% with targeted indexes.
    • Real-time file sharing to WhatsApp using Node.js streams and Cloudflare.

  5. ongoing

    Cyos Technologies

    Backend Software EngineerSelf-employed · RemoteBrazil

    • Freelance backends with AI-powered APIs for mobile apps, and RESTful APIs shared across web and mobile.
    • Schema restructuring and query optimization for better reliability and less downtime.

Universidade Paranaense (Unipar)

Associate degree in Systems Analysis and Development

Stack

What I reach for, grouped by layer.

AI & agents
  • Agno
  • RAG
  • pgvector
  • OpenAI
  • Claude
  • Langfuse
  • PII & prompt-injection guardrails
Frontend
  • React
  • Next.js
  • TypeScript
  • TanStack Query & Router
  • Tailwind
  • shadcn/ui
Backend
  • Node.js
  • NestJS
  • Python
  • FastAPI
  • tRPC
  • PostgreSQL
  • Drizzle
  • Convex
Cloud & delivery
  • AWS Lambda, SQS, S3, CDK
  • Railway
  • Vercel
  • Docker
  • GitHub Actions
Mobile
  • Swift
  • SwiftUI
  • React Native
  • App Store releases

I had the pleasure of working closely with Luiz at RENDAIR, and throughout that time he consistently stood out for his professionalism, positive attitude, and strong sense of ownership.

Quim CivitArchitect & Co-founder, RendairRecommendation on LinkedIn