Database and search engineer at an Athens travel company, building hotel search with pgvector and Elasticsearch. Replaced keyword search with vector similarity search, increasing booking conversion by 22%.
Traditional anomaly detection requires hand-tuned thresholds and fails to understand context. LLMs can reason about why a metric is anomalous and whether it matters. This guide covers integrating LLMs into time-series monitoring pipelines for smarter, context-aware alerting.
19 July 2026
Bi-encoder retrieval is fast but imprecise. Cross-encoder reranking fixes the precision problem by scoring each retrieved document against the query in full context. This guide covers building a two-stage retrieval pipeline with LLM-powered reranking for dramatically better search results.
19 July 2026
A thorough guide to Anthropic's prompt caching feature — how it works internally, when the economics make sense, how to structure requests to maximise cache hits, and common pitfalls that silently disable caching
15 July 2026