Bob Treacy started his career as a union steward on the factory floor at GE Aircraft Engines. After earning a BS and MS in Computer Science from Boston University while also raising a family, he jumped into software, never looked back, and remains at the leading edge of Java and AI today. He picked up Java in the summer of 1995 at a training session in New York, and the language has been central to his work ever since. Today he is Principal Software Architect and Data Engineer at Harvard’s Institute for Quantitative Social Science, and he has attended more than 20 JavaOne conferences, which is most of Java’s lifespan.
Getting to all those JavaOne events has always been non-negotiable. “Every job I’ve ever gone to, one of the things I demand is that I be able to go to conferences like JavaOne.” The reason, he says, is simple: “It brings a lot of different perspectives, sharing experiences, learning from each other, and just a great group of people.”
At JavaOne 2026 he’ll present work from Harvard’s Dataverse project, an open source platform for preserving and sharing research data. The problem he is solving is straightforward but persistent because researchers treat the repository like a simple archive and skip the metadata that makes their datasets discoverable. His solution uses LLM embeddings and a graph database to compare new dataset descriptions against existing ones and recommend subject categories automatically.
The conversation goes well beyond his upcoming JavaOne session itself. Bob lived Java’s entire evolution from the rough early days of applets through HotSpot, modularity, and the modern work on GPU access via Project Babylon and the Foreign Function and Memory API. On AI, he says that “I think there has to be a human in the loop.” He sees LLMs as causal models, not reasoning systems. “Given these words that came before, what’s the next word going to be? That’s not really intelligence. I don’t totally buy some of the narratives that are out there about AGI.” His advice to students is equally grounded in simplicity. He advises students to learn the code, but spend more time understanding how large systems work together. As he puts it, “just doing pure coding” is no longer enough.