Analytics Engineer. I build data pipelines and models with Databricks, PySpark, SQL and AWS, so Finance and Operations teams get numbers they can trust in Power BI.
The medallion architecture I use: records move from raw to cleaned to business-ready.Move your pointer over the raw zone
Based in
Brisbane, Queensland
Experience
6 years in data roles
Australian experience
World Courier, Brisbane
Industries
Logistics, banking, construction, brewing
Certified
4× Microsoft, incl. DP-600 & PL-300
About me
I'm Kevin, a data professional from Brazil, now living in Brisbane.
I started in operations, on a brewery production floor, working every day in SAP. Since then I have built dashboards, ETL pipelines and Lakehouse platforms for banking, construction and manufacturing companies. In Brisbane, I supported the office of World Courier, a global logistics company, with data projects.
I like work where the data is messy and the business question is clear. My job is to connect the two.
How I work
Three layers between a messy source and a clear answer.
I started on a brewery production floor, counting stock against SAP. I know what bad data costs people on the ground. So every project follows the same path.
Bronze · Land
Keep the raw data, untouched.
Ingest every source as it arrives, so any number can be traced back to where it came from.
Pulled production data from SAP PP across 7 manufacturing plants.
Silver · Clean
Fix, join and test it.
Remove duplicates, align keys and types, and add quality checks before anyone builds a report on top.
Built ETL in Pentaho and PySpark to make data consistent across sites.
Gold · Model
Shape it for the decision.
Model fact and dimension tables around real business questions, then tune them for fast dashboards.
Delivered Power BI models for Finance, Commercial and Operations teams.
Selected work
Nine selected projects, from Australia and Brazil.
I have built many more dashboards and data pipelines. These are some I chose to show. Open any project to read the full story in English, with the live dashboard. All numbers are changed to protect confidential data. Ask me if you would like to see more.
AUAustralia · Brisbane
World Courier (Cencora) · 2026
Comparing the cost of two third-party carriers
Spend, cost per shipment and cost per km for two contract carriers, with route maps across South East Queensland and Northern NSW.
From the plant floor to the Lakehouse, and now Brisbane.
Every role added a layer: operations first, then reporting, then the platform underneath.
May 2026 – Aug 2026
Warehouse Assistant, Data Projects
World Courier (Cencora) · Brisbane, QLD · Contract
Supported the Brisbane office of a global specialty logistics company. Built Power BI reports on packaging use and third-party carrier costs, alongside daily warehouse work.
Jan 2026 – May 2026
Analytics Engineer
Ilegra, for Tenda (construction & real estate) · Remote
Led the move of a legacy data platform to a Databricks and AWS Lakehouse with Medallion layers in Delta Lake.
Mar 2025 – Dec 2025
Data Analyst
Ilegra, for Banco BV (banking) · Remote
BigQuery datasets, Power BI dashboards, A/B tests and insight reports for senior stakeholders.
Jul 2023 – Mar 2025
Corporate Data Analyst
Grupo Petrópolis · Boituva, SP, Brazil
Corporate KPIs for 7 plants from SAP PP, a water consumption indicator, and Pentaho ETL.
Sep 2023 – Dec 2023
Power BI & Excel Instructor
SENAC Bahia
Taught Power BI and data analytics to working professionals, from beginner to advanced.
Nov 2022 – Jul 2023
Power Platform Developer
Grupo Petrópolis · Alagoinhas, Bahia, Brazil
ETL into the corporate data warehouse, a fleet app in PowerApps, and Power Automate alerts.
Nov 2020 – Nov 2022
MIS Assistant
Grupo Petrópolis · Alagoinhas, Bahia, Brazil
Daily work in SAP PP: production orders, materials and stock. Matched system records to physical counts and automated routines with VBA.
World Courier moves time-critical and temperature-controlled shipments. The Brisbane office uses two third-party carriers, called Phoenix and Yellow in the report. The team wanted to know which carrier gives better value, and on which routes.
All numbers in this dashboard are changed to protect privacy and confidential data. The design and logic are my real work.
Client
World Courier (Cencora)
Context
Brisbane, QLD
When
2026
My role
Warehouse Assistant with a focus on data projects
Tools
Power BIDAXPower QueryMap visuals
The problem
Carrier costs were spread across invoices and shipment records. Nobody could quickly answer simple questions: How much do we spend with each carrier? What does one shipment cost? Which routes are expensive per kilometre?
What I built
Joined shipment records with carrier costs and route distances in one model.
Created the main measures in DAX: total spend, shipments, average cost per shipment and AUD per km.
Built six report pages: Overview, Map, Carrier Comparison, Route Detail, History and Map Detail.
Added map views of each route, from Brisbane to the Gold Coast, Sunshine Coast, Toowoomba, Wide Bay and Northern NSW.
Added filters for carrier, origin suburb, destination suburb and date, plus an average line on daily spend.
Key measures
Total Spend
All carrier charges in the selected period, in AUD.
Average cost per shipment
Total spend divided by the number of shipments.
AUD per km
Total spend divided by total kilometres. This lets you compare short and long routes fairly.
Spend by day
Daily spend for each carrier, with the period average as a reference line.
Map Detail page: routes on the map and a table of cost per km and average cost per shipment for each origin and destination region.
Tracking packaging use for temperature-controlled shipments
Medicines and clinical trial samples travel in special packaging: dry ice boxes, dry shippers, pallets and insulated boxes. Each type keeps a different temperature. This report shows how the Brisbane office uses that packaging every day.
All numbers in this dashboard are changed to protect privacy and confidential data. The design and logic are my real work.
Client
World Courier (Cencora)
Context
Brisbane, QLD
When
2026
My role
Warehouse Assistant with a focus on data projects
Tools
Power BIDAXPower QueryForecasting
The problem
Packaging is expensive, and running out of the right box can delay a shipment. The team needed a clear view of which box families they use, for which temperature bands, and how use changes over time.
What I built
Modelled box usage by box family, box size, temperature band and date.
Built KPI cards for total boxes used, total shipments, voided rate and share of non-World Courier boxes.
Showed use by box family and by temperature band (refrigerated, frozen and ambient).
Added a daily trend with a short forecast, to help plan stock.
Built a separate Stock page and a simple start page with navigation.
Key measures
Total Boxes Used
Number of boxes used in shipments in the selected period.
Voided Rate
Share of boxes that were voided, for example when a shipment was cancelled or repacked.
% N-WC Box
Share of boxes that are not World Courier's own packaging (the "Box Non-WC" family).
Temperature band
Refrigerated, frozen or ambient. Each band needs different packaging.
Start page of the report, with navigation to Overview and Stock.
Tenda is one of Brazil's largest home builders. Its data platform had grown over many years, and reports were slow and hard to trust. I worked on the team that moved it to a modern Lakehouse. It was a very large project, with data from finance, sales and construction operations.
I can't show the real system or data, because they belong to the client. The diagram and code below are examples I made for this portfolio to show how this kind of platform works.
Architecture illustration, made for this portfolio.
Client
Tenda via Ilegra
Context
Construction & real estate · Remote
When
2026
My role
Analytics Engineer
Tools
DatabricksPySparkSpark SQLDelta LakeAWS S3Power BI
The problem
Data lived in many legacy systems. Each report rebuilt the same logic in its own way, so numbers did not match between teams. Pipelines were slow, and it was hard to know where a number came from.
What I built
Designed and built distributed pipelines in PySpark and Spark SQL for financial and operational data.
Set up the Medallion layers in Delta Lake: Bronze for raw data, Silver for cleaned data, Gold for business models.
Modelled analytical datasets (facts and dimensions) for Finance, Commercial and Operations.
Connected Databricks to Power BI and tuned the datasets so reports load faster.
Added quality rules between layers, so problems are caught before they reach a report.
Example code
A short PySpark example of the kind of logic I wrote. It is not the client's code.
# Example only: Silver layer job, written for this portfoliofrom pyspark.sql import functions as F
bronze = spark.read.table("bronze.finance_invoices")
silver = (
bronze
.withColumn("invoice_date", F.to_date("invoice_date", "yyyy-MM-dd"))
.withColumn("amount", F.col("amount").cast("decimal(18,2)"))
.filter(F.col("invoice_id").isNotNull())
.dropDuplicates(["invoice_id"])
)
# Simple quality gate before the table is publishedassert silver.filter(F.col("amount") < 0).count() == 0
silver.write.format("delta").mode("overwrite").saveAsTable("silver.finance_invoices")
Analytics for a bank's Finance and Operations teams
Banco BV is one of Brazil's biggest banks for vehicle finance. I worked with its Finance and Operations teams, who needed reliable data and clear reports from very large volumes of transactions.
I can't show the bank's real dashboards or data. The diagram and SQL below are examples I made for this portfolio to show the type of work.
Architecture illustration, made for this portfolio.
Client
Banco BV via Ilegra
Context
Banking · Remote
When
2025
My role
Data Analyst
Tools
Google BigQuerySQLPythonPower BIPower AppsPower Automate
The problem
Leaders needed one trusted view of key numbers. Some processes and reports were not reliable, and the bank's internal data school tracked certificates by hand.
What I built
Built analytical datasets in Google BigQuery and connected them to Power BI dashboards.
Analysed large transaction data with SQL and Python, using statistics to find useful insights.
Ran A/B tests and backtests to check changes before rollout, improving reliability and report accuracy.
Built a Power Platform app to manage certificates for the bank's data school, replacing manual tracking.
Wrote regular insight reports for senior stakeholders on the data maturity programme.
Example code
A short BigQuery SQL example of the kind of logic I wrote. It is not the client's code.
-- Example only: compare a test group with a control groupSELECT
test_group,
COUNT(*) AS customers,
AVG(processing_hours) AS avg_processing_hours,
COUNTIF(report_error) / COUNT(*) AS error_rate
FROM analytics.process_experiment
WHERE run_date BETWEEN'2025-06-01'AND'2025-08-31'GROUP BY test_group;
Grupo Petrópolis is one of Brazil's largest brewing groups. Breweries use a lot of water, so the company wanted one way to measure water efficiency in all 7 plants.
Open the live dashboardOpens Power BI in a new tab. The report labels are in Portuguese; the translation guide is below.
All numbers in this dashboard are changed to protect privacy and confidential data. The design and logic are my real work.
Client
Grupo Petrópolis
Context
Brewing · 7 plants across Brazil
When
2023–2025
My role
Corporate Data Analyst
Tools
Power BIDAXSAP PPPentaho
The problem
Each plant measured water in its own way, so plants could not be compared. Leaders also needed the numbers for sustainability reporting.
What I built
Led the design of a water index: water used for each unit of drink produced.
Split the index by area: packaging, brewing process, and syrup room plus PET bottling.
Compared each plant's actual index with its target, with green and red status icons.
Compared actual water use with the plan, and showed production volume next to it.
Added a map of Brazil sized by consumption, and filters for date, plant, cost centre and type.
Key measures
Global water index
Water used divided by production, for all plants. Lower is better.
Beer / non-beer index
The same index, split into beer and other drinks.
Actual vs target
Each plant's index against its goal. Green means on target.
% of plan
Actual water use divided by planned water use.
Translation guide
The dashboard was built in Portuguese. Use this table to read the screen.
A corporate quality dashboard, from SAP to Power BI
ICP is the group's corporate quality indicator for brewing. It shows what share of quality checks meet the standard, from the brewing process to packaging and the finished product. I built this project end to end: the data pipeline and the dashboard.
Open the live dashboardOpens Power BI in a new tab. The report labels are in Portuguese; the translation guide is below.
All numbers in this dashboard are changed to protect privacy and confidential data. The design and logic are my real work.
Client
Grupo Petrópolis
Context
Brewing · quality across all plants
When
2024–2025
My role
Corporate Data Analyst
Tools
SAPOracleScheduled loadsPower BIDAX
The problem
Quality results lived in SAP tables at each plant. There was no automatic way to bring them together, so the corporate team could not see quickly which plant, product or process step was pulling quality down.
What I built
Designed the data capture from SAP tables into an Oracle database.
Set up scheduled loads, so the data refreshes automatically. The start page shows the date and time of the last update.
Modelled the ICP by plant, wort type (Pilsen or pure malt), area and process step.
Built a decomposition tree to drill down from the overall ICP to one step, for example maturation in one plant.
Added pages for overview, analysis by area, non-conformities and detail, with filters for year, month and plant.
Key measures
ICP
Share of quality checks that meet the standard. Higher is better.
Process / Packaging / Finished product
The same index for each big area of the brewery.
Non-conformity
A quality check that failed, with its area and step.
Start page with navigation and the time of the last data update.
Translation guide
The dashboard was built in Portuguese. Use this table to read the screen.
Bottled production: actual vs plan, by plant and line
Packaging is the last step in a brewery: beer and soft drinks go into cans, glass bottles, kegs and PET bottles. This dashboard shows how much each plant packaged, and how that compares with the plan.
Open the live dashboardOpens Power BI in a new tab. The report labels are in Portuguese; the translation guide is below.
All numbers in this dashboard are changed to protect privacy and confidential data. The design and logic are my real work.
Client
Grupo Petrópolis
Context
Brewing · 8 plants
When
2024
My role
Corporate Data Analyst
Tools
SAPOracleScheduled loadsPower BIDAX
The problem
Production planning needed one daily view of packaged volume across all plants and lines, built on the same automated SAP data flow as the quality dashboard.
What I built
Reused the SAP to Oracle pipeline with scheduled loads, so the report updates on its own.
Showed packaged volume in hectolitres by drink type, beer style and packaging type.
Compared actual and planned volume for each plant, and showed volume by production line.
Added pages for planning, projection and other indicators, with filters for plant, year, month and day.
Key measures
Actual volume (HL)
Hectolitres packaged in the period. 1 HL = 100 litres.
Planned volume (HL)
The production plan for the same period.
% actual vs plan
Actual volume divided by planned volume, for each plant.
Start page with navigation and the time of the last update.
Translation guide
The dashboard was built in Portuguese. Use this table to read the screen.
A drinks company delivers to thousands of shops and bars. This report shows how well those deliveries perform, and why some orders go wrong.
Open the live dashboardOpens Power BI in a new tab. The report labels are in Portuguese; the translation guide is below.
All numbers in this dashboard are changed to protect privacy and confidential data. The design and logic are my real work.
Client
Grupo Petrópolis
Context
Logistics
When
2017–2021 data
My role
Power Platform Developer / Data Analyst
Tools
Power BIDAXPower QueryOracle SQL
The problem
The logistics team had delivery data but no single view of service level. They needed to see OTIF over time and find the main reasons for failed or late orders.
What I built
Calculated OTIF (On Time In Full), On Time and In Full rates.
Showed order volume and order cycle time, with trends by year and quarter.
Compared on-time and late orders for each year.
Ranked delivery problems by reason (for example missing order, wrong price, shop closed) and by branch.
Mapped orders by state and city, with filters for year, month and branch.
Key measures
OTIF
Share of orders delivered on time and complete. The main service level KPI.
On Time
Share of orders delivered by the promised date.
In Full
Share of orders delivered with all items.
Order cycle time
Average days from order to delivery.
Start page: "Logistics and data analysis. In this report you can see the main logistics KPIs and analyse them quickly and easily."
Translation guide
The dashboard was built in Portuguese. Use this table to read the screen.
