Dennis Oleksyuk - Inside building real AI agents for the messy world of air-freight logistics.

On this episode of Data Day, Greg Michaelson sits down with longtime friend and former DataRobot colleague, Dennis Oleksyuk, co-founder and CTO of AirCon, to unpack what it really takes to put AI agents into production in a weird but massive industry: air freight.Dennis shares his unconventional path from growing up in one of the coldest inhabited places on earth, to applied math in a Soviet-style university, to telecom engineering, and eventually into machine learning and AI infrastructure. He explains how AirCon’s agents read freight emails, build viable multi-leg routes, talk to a zoo of carrier APIs, and autonomously generate and book quotes for freight forwarders, all while dealing with a million messy corner cases that never appear on the public internet.Greg and Dennis dig into the reality of agentic coding in production: why generic agent frameworks often fall apart, why software engineering fundamentals matter more than “prompt wizardry,” how context windows really work, and why fine-tuning is usually limited by training data, not tooling. Along the way they touch on hidden operational knowledge in every industry, the myth of microservices as a default, and Dennis’s infamous “get on a plane, B” bug story.If you care about AI agents, real-world automation, infrastructure, or just want a peek behind the curtain of how your stuff actually gets around the world, this one is for you.