The Agent Harness: Where Reliability Actually Lives
The model is a dependency you will swap twice a year; the harness around it is the product. A redesigned architecture, stage by stage, and an honest guess at where failures come from.
I help teams size Redis properly
Senior Solution Architect at Redis, covering India and South East Asia. Ten years writing Java backends before this, which is why I'd rather size your cluster and run the benchmark than walk you through a slide.
Recent writing
Vector Sets vs the Query Engine: Picking the Right Primitive
Redis has two ways to search vectors and the docs never put them side by side. One is a data type, the other a search engine, and cluster mode settles it more often than latency does.
Where RDI Fits in Media and Entertainment (and Where It Doesn't)
Catalogue, rights windows and entitlements are a near-perfect fit. Watch progress, session limits and live feeds are not, and the two look alike on a whiteboard.
Semantic Caching: Cutting LLM Cost Without Cutting Corners
Break-even sits near a 1% hit rate, so the cost case is easy. The threshold, the tenant boundary and what you refuse to cache are where it actually gets decided.
Keeping Redis Honest Against a System of Record: RDI Architecture and When to Use It
How RDI is actually built, from Debezium through to the three planes, and the specific workloads where it is the wrong answer. Including the throughput numbers worth memorising.