I am Jason Stubblefield. I have spent twenty years making search engines find the right answer.

My career spans the full search stack, from racking infrastructure and tuning JVMs to setting relevance strategy and leading engineering teams. I have run a 600-million-document Solr Cloud estate on hundreds of servers, unified discovery across the library systems of a world-class business school, and most recently designed the next-generation search platform for a national e-commerce marketplace, where a controlled A/B test against the legacy system showed a sitewide sales lift of more than ten percent.

That last number matters to me because it captures how I think about search: it is not an infrastructure cost center, it is a product with measurable business outcomes. Ranking changes should be experiments, relevance should be evaluated with data rather than opinion, and the distance between an architecture diagram and a revenue number should be as short as possible.

The most interesting work happening in search right now sits at the intersection of classic information retrieval and large language models. I have shipped vectorized image retrieval over a catalog of millions of designs, and built RAG pipelines that ground an LLM in live index and query data so it can evaluate and tune ranking behavior. My consistent finding is that the two disciplines need each other: LLMs make retrieval dramatically more useful, and precise retrieval is what keeps LLMs honest.

I also build and lead teams. I have hired, mentored, and managed search engineers across three organizations, including onshore and offshore teams and a dedicated search department. I care about growing engineers into people who can own a system end to end, and about leaving every team with better operational habits than I found.

Outside of client work, I run self-directed research platforms, Recipe Saint and Room Saint, which I built end to end to test production AI search ideas: hybrid lexical and vector retrieval over millions of documents, RAG for query understanding, and Model Context Protocol integrations that let AI assistants use them directly. They are where I try things before I recommend them.

Before software, I spent years in hospitality and international business, and I still cook seriously and shoot fine art photography. That background shapes how I work: search, like a good restaurant, succeeds or fails on how it makes people feel in the first few seconds.