Available for full-time or contract roles

I build whole products,from the data pipelineto the pixels,and keep them running.

Luis Nicolau · Senior Full-Stack Engineer · AI Products

12+ years shipping products end to end in TypeScript, React, Node.js and Python, focused lately on LLM/RAG features in production. Previously Tech Lead, and co-founder/CTO of an edtech with 100k+ users.

Remote · Brazil (UTC−4) · full US overlap

Get in touch Download CV

Ask about my work

Answered by an LLM (Claude) from this site's content only, with numbered sources. It can still slip; the pages are the record.

How this chat works

Your question is checked first: length, injection attempts, bots (Vercel BotID) and per-visitor rate limits.
Hybrid retrieval finds the passages: BM25 keyword search and multilingual embeddings over this site's content, merged with reciprocal rank fusion.
Claude Haiku answers from those passages only, cites them, and refuses anything off-topic.
The answer streams back; a canary check stops it if the instructions ever leak. If the AI is unavailable, the instant keyword answers take over.

years shipping software: 12+
products founded and shipped: 4
processed by my payments startup in year one: R$5M+
AI analyses from my Claude pipeline: 29k+
users at an edtech I co-founded: 100k+

01About

Engineer, founder, tech lead.

Luis Nicolau, smiling, in a black T-shirt against a gray background

Based in: Campo Grande, Brazil
Time zone: UTC−4 · full US overlap
Experience: 12+ years, since 2014
Languages: Portuguese, English, Spanish
Open to: Full-time or contract

I'm a full-stack and data engineer with 12+ years of building products end to end, most of them as a founder or tech lead who owned the architecture.

Most recently I designed, built and run OlhoNaLei, a nonprofit civic-tech platform that tracks Brazil's federal legislation: 167,000+ bills ingested from the Chamber and Senate open-data APIs, hybrid semantic search on PostgreSQL + pgvector that I took from ~1.85 s to ~8 ms, and a Claude-powered analysis pipeline that has written 29,000+ plain-language summaries under a monthly budget cap. Before that I built SenseComment, a React/Node SaaS on top of the YouTube APIs that grew to 500+ paying subscribers, and EcoDoctor, a healthcare-payments platform that processed R$5M+ in its first year. Earlier I was co-founder and CTO of Jovens Gênios, an edtech that passed 100,000 users, Tech Lead at Matrix Energia, and a contractor for The Contract Network, a US company building an LLM-based contract-analysis product.

I'm used to owning a product with little guidance: taking a vague, one-line goal and making the dozens of small decisions it takes to ship something reliable, while sweating the edge cases. I design for retries, idempotency and failure containment, and I like chasing performance problems with real measurements.

I work with AI coding agents every day (Claude Code, Cursor, Codex). I write the spec and acceptance criteria, let the agent draft, then review its plan and its diff, and catch the places where it is confidently wrong. Lately I run several agents in parallel on separate git worktrees, coordinated through a shared long-term memory I built myself.

02Selected work

Products I built, shipped and kept running.

Whole products, from the data pipeline to the pixels. Each one has a case study with the problems, the decisions and the numbers.

FlagshipLive

OlhoNaLei

Founder & sole engineer · 2026 – Present

Nonprofit civic-tech platform that makes Brazil's federal legislation understandable.

bills ingested: 167k+
AI analyses written: 29k+
faster vector search: ~220×
missing historical bills recovered: 8,300+

Read the case study Live site

OlhoNaLei page for the tax-reform bill PLP 68/2024, with its status, its path through Congress and its plain-language summary.

sensecomment.com home page: “Stop toxic comments before they hurt”, AI-powered YouTube comment analysis.

Live

Product · 2025 – Present

SenseComment

Founder & engineer

AI-powered YouTube comment analysis that filters toxicity and spam and finds leads; 500+ paying subscribers.

React
TypeScript
Tailwind CSS
Node.js

Read the case studyLive site

Paneleiro App website: “Aqui você joga junto”, with App Store and Google Play buttons.

Live

Product · 2025 – 2026

Paneleiro

Founding Engineer

Live football fan app for iPhone and Android: real-time matches, push notifications and highlights.

C#
.NET 8
EF Core
Hangfire

Read the case studyLive site

EcoDoctor

R$5M+

in transactions in year one

Product · 2023 – 2025

EcoDoctor

Founder & lead engineer

Healthcare-payments platform that processed R$5M+ in transactions in its first year.

Node.js
TypeScript
MySQL
MariaDB

Engineer, end to end

Online booking for beauty salons, with an AI attendant on WhatsApp.

Next.js
React 19
TypeScript
Tailwind CSS

DetailsLive demo

04Experience

Jobs, and the products I ran alongside them.

Overlapping dates are real: I often ran a product of my own next to a job.

OlhoNaLei · Founder & sole engineer · 2026 – Present
SenseComment · Founder & engineer · 2025 – Present
Paneleiro · Founding Engineer · 2025 – 2026
EcoDoctor · Founder & lead engineer · 2023 – 2025
The Contract Network · Senior Full-Stack Engineer · 2023 – 2025
Matrix Energia · Tech Lead · 2022 – 2024
Jovens Gênios · Co-founder & CTO · 2017 – 2022
Wikki Brasil · Software Engineer · 2017 – 2018
Caronaê · Android & Full Stack · 2015 – 2017
Fluxo Consultoria · Full Stack Web · 2014 – 2016

Employment

The Contract Network
2023 – 2025
Senior Full-Stack Engineer (AI Applications)
Contract via Deel · remote · US company (Georgia)
- Built serverless, event-driven backend services on AWS for an LLM-based contract-analysis product: Lambda, Step Functions, EventBridge, SQS, API Gateway and MSK/Kafka.
- Orchestrated multi-step workflows with Step Functions and EventBridge, wiring the AI/ML team's model outputs into the application and its APIs.
- Wrote services in Node.js + TypeScript with Apollo GraphQL and Neo4j, and optimized the Neo4j read/write paths.
- Built the AWS infrastructure as code, with automated container deploys on ECS.
- Used Copilot and Cursor daily for code, tests and docs, with unit and integration tests validating generated code before merge.
Matrix Energia
2022 – 2024
Tech Lead / Full-Stack Engineer
São Paulo, Brazil · energy
- Led a full-stack team building service-based systems (Node.js services, REST APIs, messaging) integrated with AWS and Azure.
- Introduced CI, code review and testing standards that reduced production incidents by ~30%.
- Started the team on AI-assisted coding (Copilot, ChatGPT) in late 2022, with human review and automated tests as guardrails.
Jovens Gênios
2017 – 2022
Co-founder & CTO
Rio de Janeiro, Brazil · edtech
- Co-founded the company and led engineering for edtech apps that grew past 100,000 users on web and mobile.
- Chose the stack and the delivery process; scaled and evolved the PHP (Laravel) backend in production, and moved the data architecture from flat structures to a relational model.
- Helped build machine-learning models that recommended content to students and predicted their performance with Item Response Theory (IRT); my main part was deploying them to production.
- The company reached a ~R$100M valuation.

Earlier · 2014 – 2018

Wikki Brasil · Software Engineer / Full Stack
2017 – 2018
Web systems with REST APIs, across the full development cycle.
Caronaê · Android & Full Stack Developer
2015 – 2017
A university ride-sharing Android app (UFRJ) and its backend APIs.
Fluxo Consultoria · Full Stack Web Developer
2014 – 2016
Web systems and REST APIs for a range of clients, at UFRJ's engineering junior enterprise.

Products I built end to end

05How I work

Principles I actually follow.

01
Own it end to end
From data model to deploy to support. Comfortable with a one-line goal and no spec.
02
Measure, don't guess
Benchmarks before and after. Profile the real path, network included.
03
Design for failure
Idempotent jobs, retries with backoff, flags that survive failures, canaries, guards against orphaned jobs, explicit degraded modes and rollback switches.
04
Cost is a feature
Budget caps on every LLM call path, batch APIs, runaway invocations killed, indexes sized to fit.
05
Write decisions down
ADRs, requirement specs, runbooks, and a regression test for every production bug.
06
CI that proves itself
A new workflow must run once and pass, or it gets disabled.
07
Secrets stay out of chat and code
Agents ask for them through a vault flow and never read them.
“Performance measured, not guessed.”

Working with AI agents

Every day, with judgment: the agent drafts, the tests decide.

1Spec
I write the spec and the acceptance criteria.
2Draft
The agent drafts, following each repo's instructions and skills from my template.
3Review
I review its plan and its diff, and catch confidently wrong framework or architecture calls before merge.
4Verify
Tests and checks decide when the work is done, not the agent's own claim.
5Scale
A manager agent runs workers in parallel, one git worktree each. Overnight runs log every decision for my morning review.

Claude Code every day, plus Cursor and Codex; Copilot and ChatGPT since late 2022. MCP servers for shared memory (Lore), Notion, Railway, AWS and Figma.

06Skills

What I've used in production.

Core stack

Node.js~10 yrs
Express, tRPC, GraphQL, serverless
SQL10+ yrs
PostgreSQL, MySQL, SQL Server
React~8 yrs
including Next.js 15 / React 19 in production
TypeScript~7 yrs
front to back
Python~5 yrs
FastAPI services, data pipelines
AWS serverless~5 yrs
Lambda, SQS; Step Functions, EventBridge, API Gateway
C# / .NET 8since 2025
Web API for Paneleiro

Languages

TypeScript
JavaScript
Python
C# (.NET 8)
SQL
Java (Spring Boot)
PHP (Laravel)
Dart (Flutter)
Go
Rust
Kotlin/Java for Android

Frontend & mobile

React 17–19
Next.js (Server Components, ISR, OG images)
Tailwind CSS
Vite
Vue.js 3 / Nuxt
Flutter (iOS/Android)
Capacitor
React Native
Native Android
Admin consoles
Embedded widgets

Backend

Node.js + Express
tRPC
Apollo GraphQL
FastAPI
Django
Flask
.NET 8 Web API (EF Core, Hangfire, SignalR)
Spring Boot
Go
REST & GraphQL API design

AWS

Lambda
Step Functions
EventBridge
SQS
API Gateway
ECS
EC2
S3
MSK/Kafka
CloudWatch
Infrastructure as code

GCP, hosting & ops

Cloud Run
Cloud Functions
Pub/Sub
Firebase
Vercel
Railway
Supabase
Neon
Cloudflare (R2, Workers, DNS)
Docker
GitHub Actions
Sentry

Data

PostgreSQL (tuning, full-text, RLS)
pgvector / HNSW
Binary quantization + reranking
MySQL / MariaDB
SQL Server
Redis
Neo4j
Firestore
MongoDB
Qdrant
SQLite

Distributed systems & reliability

Queues & pub/sub
Scheduled & recurring jobs
Idempotent, race-aware design
Retries & backoff
Data-freshness canaries
Orphaned-job guards
RabbitMQ
Kafka

AI / LLM

Claude API (Message Batches)
OpenAI API
Gemini
Voyage embeddings
RAG
Semantic & hybrid search
LangChain
Tool calling & structured outputs
LLM cost governance
MCP servers

AI-native development

Claude Code
Cursor
Codex
GitHub Copilot
Per-repo agent instructions & skills
Parallel agents on git worktrees
Autonomous runs with logged decisions

Ways of working

End-to-end ownership
Comfort with ambiguity
ADRs & written specs
Unit, integration & regression tests
Code review
Payments & third-party integrations
Security (RLS, LGPD, passkeys)

07Background

Education & languages

Education

Universidade Federal do Rio de Janeiro (UFRJ)

Electronic & Computer Engineering (incomplete), 2012 – 2017

Left to co-found Jovens Gênios and build software full-time.

Certifications

Harvard CS50
Udacity Android Nanodegree
Google Project Management

Languages

Portuguese · native
English · advanced
Spanish · intermediate

I also teach IT and AI in Portuguese on my YouTube channel, Seu Nicolau.

08Contact

Let's build something.

Available for full-time or contract roles. Remote from Brazil, with full overlap with US business hours.

Email meWhatsApp LinkedIn GitHub

Campo Grande, Mato Grosso do Sul, Brazil

Local time in Campo Grande

CV · English (PDF)Currículo · Português (PDF)

I build whole products,from the data pipelineto the pixels,and keep them running.

OlhoNaLei

Employment

The Contract Network

Matrix Energia

Jovens Gênios

Earlier · 2014 – 2018

Products I built end to end

OlhoNaLei

SenseComment

Paneleiro

EcoDoctor

Own it end to end

Measure, don't guess

Design for failure

Cost is a feature

Write decisions down

CI that proves itself

Secrets stay out of chat and code

Working with AI agents

Core stack

Languages

Frontend & mobile

Backend

AWS

GCP, hosting & ops

Data

Distributed systems & reliability

AI / LLM

AI-native development

Ways of working

Education

Certifications

Languages

Let's build something.