Two months building an investment bot. What it taught me about LLMs
How a portfolio-monitoring bot taught me how LLMs really work: zero-temperature non-determinism, determinism in the data, and Condorcet-style ensemble voting.
My articles on data engineering, cloud and software architecture.
How a portfolio-monitoring bot taught me how LLMs really work: zero-temperature non-determinism, determinism in the data, and Condorcet-style ensemble voting.
An end-to-end streaming pipeline on the Olist dataset: revenue aggregation, CEP anomaly detection, per-minute KPIs, Iceberg data lake on MinIO — and the problems you don't see coming.
Why and how I designed a lakehouse with no vendor lock-in: Kafka, Spark, Iceberg, MinIO, Great Expectations and Terraform.
How I designed a pipeline of 4 specialized LLM agents with LangGraph, prompt caching and automatic fix loop.
What I learned stepping away from code to manage sales data production — alerting, postmortems, crisis communication, and why I went back to engineering.