Andrew Kornilov·4d agoCan Decision AI pick a product’s main catalog image? And what does it cost?The task
Andrew Kornilov·Sep 15Teaching a small model to judge like AmazonIt seems pretty straightforward: use a large LLM to judge your search results — the bigger LLM, the better. But at scale, this approach is…
Andrew Kornilov·May 29The Problem Was Never the ContentBack in October 2019, I escaped rainy Amsterdam for a few days and flew to sunny Valencia. One of those small trips that somehow reset your…
Andrew Kornilov·Apr 27Building a Provider-Agnostic LLM Batch Pipeline in DjangoHow We Wired Mistral and OpenAI Batch APIs Into a Reusable Processing Engine — and What We Learned the Hard Way
Andrew Kornilov·Apr 10How to Evaluate Image Search in Qdrant Using Quepid (And the Hacks It Takes). Part 2(Hacks)This article was born from an idea to share some neat hacks in Quepid, but it quickly became clear that some of those hacks were no longer…
Andrew Kornilov·Feb 15How to Evaluate Image Search in Qdrant Using Quepid (And the Hacks It Takes). Part 1 (No Hacks)Initially, this was meant to be a single article about evaluating an image search system powered by vector search in Qdrant. However, I…
Andrew Kornilov·Jan 20MicrologAI vs CeleryNothing suggested trouble that day. Then, out of nowhere, everything falls apart. We’ve all been in situations like this. This is one of my…
Andrew Kornilov·Jan 8Mirror, mirror, what’s behind you?In the first part of my exploration of the MCP capabilities that Elasticsearch provides, we focused on the monitoring cluster. Now it’s…
Andrew Kornilov·Dec 25, 2025Robots vs zombies in twilight battleElasticsearch management for l̶a̶z̶y busy folks…