Engineering Notes
Evergreen technical references, organised by domain. These are written to be returned to: each one covers the reasoning behind a technique, the trade-offs that decide when to reach for it, and the failure modes that show up at production scale.
Many articles sit in more than one domain — machine learning applied to satellite imagery, or deep learning that only runs at all on a cluster. The navigation files each article in one place; the tag index shows the cross-links.
-
Statistical foundations, transformer architectures, LLM inference mechanics, and fine-tuning strategies for adapting foundation models to domain-specific targets.
-
The geospatial Python stack, coordinate reference transformations, state estimation and sensor fusion, and the sensor physics behind hyperspectral retrieval.
-
Distributed execution across cluster nodes, self-hosted model serving on institutional GPUs, and reproducible environment configuration for shared systems.