In the 11th episode of "Clojure in product. Would you do it again?", Artem Barmin and Vadym Kostiuk speak with Jeremiah Via, Staff Software Engineer at The New York Times. Jeremiah describes how Clojure was introduced and adopted across the search stack at a major media organization, and why JVM interop and practical tooling made it the right choice for their data-processing workloads.
Our conversation walks through concrete topics: Jeremiah’s Clojure origin story, the iterative migration from PHP, Erlang, Python, and Java to JVM/Clojure services, and the search team’s day-to-day work, including how they push vector embeddings into Elasticsearch for AI features and performance.
We also dig into hiring and engineering practices: onboarding newcomers with an emphasis on functional thinking and REPL workflows, hiring for search/domain expertise over prior Clojure experience, maintaining code discipline, and addressing production concerns like memory sizing and performance tuning. As Jeremiah notes, "Now with AI stuff, people can be productive very fast without understanding it, using a cursor and tools like that," and he cautions that it remains to be seen how this will affect the deeper mental model of learning to think in Clojure.
We conclude with Jeremiah's response to a question from our previous guest, Cam Saul from Metabase.
Listen to our podcast and get more insights about Clojure in product: https://www.freshcodeit.com/podcast/clojure-still-gives-the-biggest-performance-boost
Top comments (0)