Corporate knowledge and verifiable answers
Search across complex documents with cited evidence, controlled corpora, permissions, model cost, and answer quality.
Anonymized engineering notes
We publish only material that can be described without identifying parties or making unverified outcome claims: context, constraint, architecture decision, and tradeoff.
engineering topics
Recurring engineering problem classes
AI search, geodata, and registries need the same discipline: normalized data, clear permissions, verifiable sources, and observable operations.
Maps and territorial analytics
Maps as operational tools: layers, data provenance, object search, territorial comparison, and reliable mobile use.
Registries and selection systems
Catalogues and registries with normalized data, deadlines, sources, filters, and controlled updates.
anonymized reviews
Three engineering decision scenarios
Each note records the original constraint, selected architecture, and accepted tradeoffs. Parties, outcomes, and metrics are not disclosed.
Payments: service boundaries and resilience
How to establish service ownership, plan a safe migration, and make controlled changes to a critical system.
Sensitive-data service: access and audit logs
What to decide when access roles, audit logs, data quality, and interface resilience matter.
Catalogue: search, cache, and operations
How to separate search, caching, queues, and observability so changes remain controlled.
engineering journal
Engineering journal on decisions and cost
The journal includes articles on AI cost, monoliths, TypeScript, RAG, Kubernetes, and PostgreSQL under load.
How to reduce language-model costs
Routing, caching, batching, and verifiable economic impact.
When a monolith is better than microservices
When a monolith is cheaper, faster, and more reliable than microservices.