Conference talks

Composing async event streams from heterogeneous scientific instruments — same ReactiveX operator vocabulary across Tango, EPICS, and TINE, with a shared upstream multicast layer that keeps SCADA load flat regardless of consumer count.

Zero-native-dependency HDF5 writer for scientific data pipelines: how to hit raw write throughput targets without JNI and what that means for portability in containerized environments.

How to structure a performance assessment of a distributed metadata catalog: scenario design, load model, and where the ceiling actually sits before capacity planning.

Methodology for reproducible cross-language benchmarks: 72 scenarios across four runtimes, eliminating toolchain bias before comparing numbers across 20+ research facilities.

Infrastructure cost reduction as a platform decision — VPA + rightsizing found savings the migration itself had missed. Zero application code changes.

Running production SCADA on MicroK8s: what operator UIs need from a container runtime and how to keep the facility running during a rolling upgrade.

Applying distributed tracing to instrument control: how APM primitives designed for microservices translate to Tango device servers, and what you can see that logs alone cannot show.

Instrumented with Elastic APM before migrating, surfaced sequential Elasticsearch fan-out no application metric had caught, cut P99 33% — the migration became an observability project.