Blog
Original writing on architecture, event-driven systems, reliability and engineering careers — drawn from production work on banking and high-traffic platforms, with the numbers and the trade-offs left in.
14 articles
This site has its own AI Assistant, grounded in its own content through a real retrieval-augmented generation pipeline: embeddings, a relevance gate, access-aware retrieval, and a semantic answer cache. Here is how each piece earns its place.
How we cut L1 billing-support resolution time in half with a Python LLM service that streams answers over SSE and retrains itself on PyTorch from every query it has already resolved, sitting next to a Java and Spring Boot platform it never has to touch directly.
Architecting a corporate pricing and billing platform where the number has to be reproducible in front of an auditor a year later: why the pricing engine is a Kafka consumer instead of a service you call, how the command and query sides split, and what partitioning actually buys you.
Circuit breaker states, autoscaling on breaker signals instead of raw CPU, Thanos for metrics that survive a Prometheus restart, and why Azure Monitor sees failures Resilience4j never will — from the work that took uptime from 99.5% to 99.95%.
An architecture for an API gateway that verifies identity once, enforces rate limits atomically, and decouples every downstream consumer via async events — with the config details most write-ups skip.
Two markdown files, updated in the same commit as the code, turn an AI coding agent from guessing at your codebase's intent into actually knowing it — no vector database required. Here is the pattern, working, on this project.
