Insights
Notes from the build,
and from the pager.
Engineering writing on platform architecture, feed normalisation, regulated wagering systems and what running production around the clock actually costs. Written for the people who have to maintain it.
11 posts · page 1 of 2
- Engineering
Act before your customer notices
Detect a production fault, open the ticket with the affected customer attached, and tell that customer first. How we build it and what it takes.
Protocolzone
- Engineering
AI-generated proposals need an approval path
Inside an implemented AI workflow: structured responses, persisted evidence, human review and retry handling before a proposal reaches the platform.
Protocolzone
- Engineering
Multi-tenant wallets: a timeout is not a failed debit
An implemented wallet integration separates uncertain debits from retryable operations and limits concurrent calls per tenant to contain slow dependencies.
Protocolzone
- Engineering
What it takes to run a model in production, not just train one
The toolchain and discipline behind three shipped model families: point-in-time factors, a simulation engine, promotion gates, and retraining without regression.
Protocolzone
- Industry
Blockchain and the tote: what a provably-fair pool would take
Pari-mutuel already computes a dividend from a pool everyone pays into; the punter just never sees the pool. A design essay on closing that gap on-chain.
Protocolzone
- Industry
Where a blockchain earns its place in government systems
A design essay on where a chain earns its place in public systems: provenance registries and programmable settlement, and where it does not.
Protocolzone
- Engineering
Containers or serverless: the cloud estate we actually run
Why our multi-tenant wagering platform runs on GKE while the tipping app ran serverless on Lambda: synchronised race-day peaks versus event-driven spikes.
Protocolzone
- Engineering
Elasticsearch as a projection: global search on a wagering platform
Why we run Elasticsearch beside the ledger on a wagering platform we operate: operator search, player lookup and search-based analytics, index as a projection.
Protocolzone
- Engineering
Kafka as the ingestion spine of a racing data pipeline
How we run a 24x7 racing data pipeline on Kafka: correction-safe topic design, replayable history, and normalisation done once at ingestion.
Protocolzone