I’m a software engineer with 12+ years of experience building large-scale distributed systems and data infrastructure. I’m interested in the boundaries between dependable data systems, AI engineering, software architecture, and security—especially the details that determine how systems behave in production.
- Query engines and storage systems: correctness, pruning, execution, and observability.
- GenAI infrastructure: evaluation, telemetry, privacy, and operational failure modes.
- Security as an engineering constraint, from supply chains and deployment pipelines to the data retained for debugging.
I contribute to Apache DataFusion, a query engine and toolkit built with Rust, with recent work around Parquet correctness, pruning, and execution metrics. I’ve also contributed container and CI improvements to OWASP PwnzzAI, an intentionally vulnerable application for learning about AI security.
A reproducible study of the trade-off between forensic utility and plaintext exposure in GenAI telemetry retention. The repository includes a frozen study protocol, implementation, automated tests, privacy-safe aggregate data, and validation artifacts.
SparseVector is my independent space for practical experiments, technical explanations, and observations from building software. I write about data infrastructure, AI systems, architecture, and security, with an emphasis on mechanisms and reproducible examples.