MOAI DATA
Data engineering · Analytics · AI-ready data

Data your AI can trust.

Moai Data builds the pipelines, data models and semantic layer behind your dashboards and AI assistants. One source of truth for your business metrics.

Get in touch See success cases

Services.

Start where you are. Not sure? A data foundation audit gives you a readiness report and a prioritized plan.

01

Data engineering

Automated pipelines into one warehouse. Tested, monitored and documented.

Google BigQuery · Snowflake · Python · Prefect · Dagster · Airflow
02

Analytics engineering

dbt models, metric definitions and the dashboards your teams open every day.

dbt · Looker · Power BI · Metabase
03

AI-ready data

A semantic layer and MCP server, then assistants and agents that answer with numbers you can check.

Claude · Claude Code · dbt Semantic Layer · MCP
04

Machine learning

Forecasting, classification and advanced analytics on your own data.

Python · Forecasting · Classification

Who is this for?

Companies whose data lives in many tools, and who want to be data-driven without hiring a full data team. Mostly small and medium-sized businesses and startups.

E-commerce

Online stores and brands selling through their own shop, marketplaces and paid ads.

Shopify · Klaviyo · Meta Ads · Google Ads

SaaS and digital products

Subscription software, mobile apps and online platforms that need clear revenue, retention and usage metrics.

Stripe · App Store · Amplitude · Mixpanel · PostHog

Travel and hospitality

Hotels and travel companies with reservations and guest data in separate systems.

Booking systems · Property management

Telecom

Telecom companies that need revenue assurance and fraud control across their systems.

Revenue assurance · Fraud management

Problems Moai Data solves.

If one of these sounds familiar, there is a fix for it.

Every team reports a different revenue.
One source of truth. Each business metric defined once, in code.
Data is scattered across tools.
Every source centralized in one warehouse, updated when you need it.
Reports are built by hand in spreadsheets.
Automated dashboards for marketing, finance and operations.
Pipelines fail and nobody notices.
Tests, alerts and an AI agent that opens the fix for your review.
The AI assistant gives confident wrong answers.
A semantic layer, so it answers from governed metrics.

Why data-driven?

Five benefits, and five steps to get there.

Two panels. Benefits of data-driven decision making: make confident decisions, guard against biases, find unresolved questions, set measurable goals, improve company processes. How to make data-driven decisions: know your vision, find data sources, organize your data, perform data analysis, draw conclusions.

Methodology.

Six steps, from setting goals to asking AI about your own data.

Six steps on an arrow: set goals, mapping sources, design architecture, build solution, data reports analysis, ask AI about your own data.

Success cases.

Client names withheld. The stacks and the results are real.

Case 1 · SaaS mobile app

From manual Excel reports to hourly data.

Problem
Analysts downloaded data from each platform by hand and pasted it into one big Excel file. Keeping reports updated cost them hours.
Solution
Snowflake warehouse, Fivetran for fast ingestion, dbt models in 3 layers (raw, stage, mart) with quality checks, and Looker dashboards.
Result
Strategic decisions on fresh data within 1 month. Fivetran was later replaced by Airflow to cut costs, and new sources were added.
Architecture diagram: App Store, Google Ads, Appfigures, Facebook Ads, Mixpanel and Stripe data extracted with Fivetran, then Apache Airflow, into Snowflake, modeled with dbt and shown in Looker.
1 month
to the first data-driven decisions
Hourly
data refresh in the dashboards
2 stages
Fivetran first, Airflow later
Case 2 · AI agent in production

A self-healing agent for data pipelines.

Three ways to fix a failed pipeline. Manual: time-consuming debugging, hours later. AI-assisted: faster, but still copy and paste. Self-healing agent: a webhook wakes the agent, which reads the logs, finds the fix, opens a pull request and notifies an engineer, who reviews, approves and deploys to production within minutes.
8 → 2 hours
of pipeline maintenance per week, around 75% less
Around 70%
of recurring, fixable pipeline issues are candidates
1 human step
the engineer reviews and approves the pull request

Error → agent triggered → investigate → root cause → fix → pull request → engineer approval. Expired credentials, external API issues and business logic still go to a human.

Case 3 · Fashion e-commerce · Shopify and Klaviyo

One source of truth for a fashion brand.

Goal
A growing women's fashion brand needed one source of truth across store, ads and email data.
Solution
Shopify, Klaviyo, Facebook Ads, Google Ads, TikTok and Google Analytics loaded with Stitch into BigQuery (migrated from Redshift), modeled in dbt Cloud, served in Looker.
Result
Up-to-date CAC, ROAS and margin. The team optimized ad spend, improved influencer ROI and boosted email performance.
Architecture diagram: data sources including Shopify, Klaviyo, Amplitude and ad platforms, extracted with Stitch and Fivetran into Google BigQuery, modeled with dbt and shown in Looker dashboards.
40%
lower infrastructure cost after the move from Redshift to BigQuery
Delivered
Attribution modelCAC, ROAS, marginAd spendCustomer segments
Portrait of Arthur Duarte
About · Arthur Duarte

One engineer, the whole data lifecycle.

Working with data
12+years
Working in IT
20+years
Delivered on Toptal
3,000+hours
English, Portuguese, Spanish
3languages
PROFILE

Arthur Duarte is a Data Analytics Engineer. Through Moai Tech LLC, he helps small and medium-sized companies build modern data foundations, automate pipelines and turn multi-source data into clear insights and dashboards.

MBA in Project Management · BSc in Computer Engineering
Arthur on LinkedIn
ROLES COVERED
  1. 1Data Architecture
  2. 2Data Engineering
  3. 3Analytics Engineering
  4. 4Data Analytics
  5. 5Data Science
  6. 6AI Engineering
From data architecture defined from scratch to enterprise data teams.
TECH STACK
LanguagesSQLPython
BILookerPower BIMetabaseTableau
PipelinesdbtAirflowPrefectDagster
CloudAWSGCP
WarehousesSnowflakeBigQueryRedshiftPostgreSQL
AIClaude CodeLLM applications
Companies Arthur has worked with
Mobileum Parade Appex Webmotors Toptal Entel TIM Brasil Spin
Pricing
Standard data project, starting at $5,000.

The final price varies with the scope, the complexity, and the number of data sources and metrics.

Let's talk about your data.

Send a few lines about your data and what you want from it. You get a reply by email, and a video call is arranged at a time that suits you.

Your message is delivered by email through Web3Forms.

Or write to LinkedIn