I build the full data path from raw source to dashboard: ingesting with Python ETL scripts and Airbyte, modelling with dbt into tested dimensional warehouse layers, and releasing through version control and CI/CD, so every metric is defined once and can be trusted. I am now adding orchestration with Airflow and event-based streaming with Kafka.

Portrait of Eugine Ndire wearing glasses and a navy cardigan

Toolbox

Data Pipeline Architecture
Draw.io
Data Ingestion
ETL Python scripts, Airbyte
Data Warehousing
Snowflake, BigQuery, PostgreSQL
Data Modelling and Transformation
dbt
Data Analysis
Python (pandas and NumPy), SQL (Advanced SQL)
Data Visualization
Looker Studio
Version Control
Git & GitHub
CI/CD & Automation
GitHub Actions
Containerization
Docker
Agentic AI Engineering
n8n
Other Tools
VS Code, DBeaver
Learning
Orchestration: Airflow
Event-based Streaming: Kafka

Try the cleaning pipeline

A raw customer export with real-world problems. Run each step, or press Play all, and watch the data change: red cells have problems, green cells just changed.

Try the cleaning pipeline

A raw customer export with real-world problems. Run each step and watch the data change: red cells have problems, green cells just changed.

Customer data shown before and after cleaning

Data Analytics & Data Engineering Projects

Each project starts with a business question and ends with something a team can act on.

0public projects
0dbt models in the flagship
0automated data tests
0CI per pull request

nasa-etl-pipeline

Built a Python ETL pipeline that pulls natural event data (wildfires, severe storms, volcanoes, sea and lake ice) from NASA's EONET v3 API, flattens the nested JSON into one row per event reading, cleans and types it with pandas, and loads it into a PostgreSQL database on Aiven with a CSV backup. The load runs in a single transaction, so a failed run rolls back and the previous data stays intact. This is my upstream data ingestion project.

Python, pandas, requests, SQLAlchemy, PostgreSQL, DBeaver

safari-connect-analytics

Transformed fragmented, messy Safari Connect data into high-impact business intelligence. By architecting clean SQL models and complex CTEs, I built enterprise-grade analytical views that served as the foundation for Looker Studio dashboards. This project bridged the gap between raw operational data and executive decision-making, providing clear visibility into driver performance, route profitability, and passenger cancellation trends to optimize overall operational efficiency.

SQL, DBeaver, Looker, PostgreSQL

kenya-employee-salary-analysis

Built a People Analytics and Payroll Budget Audit framework for Savanna Tech Group (Nairobi, Kenya). Built with SQL to analyze a 420-person workforce, identify department budget anomalies, track internal salary equity metrics, and provide data-backed compensation insights directly to the CEO.

Advanced SQL, PostgreSQL, Looker, DBeaver

jcars-data-cleaning-project

Cleaned and standardised a raw 276-row, 32-column vehicle sales and logistics export for Jcars (Kenya). A dbt staging model on Snowflake trims text, turns placeholder values into NULL, standardises categories, repairs mixed date and money formats, flags outliers and casts every column to a proper data type. The result is a cleaned CSV with unique order IDs, no duplicates and a data-quality note on every row that was changed.

Snowflake, dbt, SQL

The whole process, from design to dashboard

Seven stages carry data from Data Analytics Pipeline Architecture to semantic modelling and Data Visualization, left to right. Pipeline reliability, n8n automation and GitHub version control run across every stage, shown by the dashed arrows. Press play for a tour, select any part of the diagram, or choose a source type in stage 2 to see how it is ingested.

Follow one record a denied claim and its encounter from my healthcare-analytics project (illustrative values), as it moves through the stages

    Deliverable
    Tools

    
              

    Version control and collaboration

    Every model, query, diagram and n8n workflow lives in Git. Work happens on branches and goes through pull requests, review and automated checks before it reaches main. Step through a typical change, from issue to release.

    main Issue #42 feature/42-add-dim-branch Pull request #42 Checks passed Approved merge commit tag v1.2

    
                  

    Automation with n8n

    n8n is my data quality reporting engine. I am still working with raw static CSV files, so each workflow reads a file, runs the quality checks, builds a short report and tells me whether the data can be trusted. Run it below, then try a file that breaks the workflow.

    Main workflow: csv_quality_report (select any node to see its configuration)

    Error workflow: starts automatically if any node above fails

    Execution log

    
    
            

    
              

    What n8n does for me

    • Starts the workflow by hand or on a schedule, whenever a new CSV is ready
    • Reads the static CSV and runs the data quality checks in a fixed order
    • Builds a plain-language data quality report with pass and fail counts
    • Sends the report by email or Slack, so I know if the data is trustworthy
    • Handles failure with an error workflow and an alert
    • Workflows are exported as JSON and reviewed in GitHub

    What the quality report covers

    • Missing values per column
    • Duplicate rows and duplicate order IDs
    • Data type and format problems, such as bad dates
    • Outliers flagged with the IQR rule
    • Row count checked against the source file

    Data engineering practices I build in

    A pipeline that runs once is a script. These habits make it something a team can rely on.

    
            

    What you get when you work with me

    Tools matter, but so does the person using them.

    Data-driven

    Decisions backed by evidence, with tables and dashboards people can trust.

    Solution-focused

    Ready to offer practical solutions to real business problems.

    Resilient

    I stay when things get thick and see the work through.

    Committed

    Ready to commit fully to the team and the goal.

    Collaborative

    Ready to collaborate and grow in a new work culture and environment.

    Resume: Eugine Ndire

    Download PDF

    Project Architecture: healthcare-analytics

    Healthcare-Analytics architecture: developers open a pull request, dbt CI builds and tests in a temporary BigQuery dataset, a merge triggers dbt CD to deploy the Bronze, Silver and Gold medallion layers to production, with OIDC authentication through Workload Identity Federation.