“I turn raw data into products people can actually use.”
Data engineering · MLOps · Generative AI on AWS & Databricks
Selected projects
Three projects that show how I work, from ingestion to the interface someone actually uses.
Stack
A knowledge graph of a classic video game, built from wiki pages, speedrun data, Wikidata and a game decompilation. Medallion lakehouse in PySpark and Delta Lake, PageRank analytics, and a Databricks App for statistics, SQL and pipeline monitoring.
Read the storyStack
The assistant on this site: a static front end calling API Gateway, a Lambda function and a foundation model on Amazon Bedrock. Serverless, pay-per-use and limited to questions about my profile.
Try itStack
An MLOps platform for investment decisions: Prophet and a PyTorch LSTM, Airflow, an MLflow registry, FastAPI serving, Evidently drift monitoring and automatic retraining. Live on a VPS with Docker Compose and GitHub Actions.
View on GitHubWhat I do
From raw data to useful intelligence. I design data platforms, ship machine-learning systems to production and build generative-AI features that stay grounded, measurable and affordable.
Expertise
Lakehouse architectures with Bronze → Silver → Gold layers, PySpark and Delta Lake, built to be tested, monitored and rerun.
Reproducible training, versioned features and automated deployment, so models keep working after the demo.
Serverless assistants on Amazon Bedrock with focused prompts, tight IAM and predictable costs.
Writing
Let's talk
Ask my assistant for a quick overview, or write to me directly. I usually answer within a couple of days.
Tap a hole to play a note
Off the clock
Before data pipelines, there was a kid with an ocarina. Ocarina of Time is the game I love most, so here's a small piece of it: each hole plays a note. The melodies are named after the songs from the game, rewritten as original tunes for this page.
Keyboard friendly: Tab to a hole, then press Enter. You can also use keys 1 to 6.