Background
Kreminski did his MS and PhD in Computational Media at UC Santa Cruz, finishing in 2022, after a BA in Interactive Entertainment at USC. He was an assistant professor at Santa Clara University (2022–2024) while joining Midjourney as Research Scientist and Director of its Storytelling Lab (2023–2025).
He moved into academia full time: Visiting Assistant Professor at Cornell Tech from July 2025, then Assistant Professor of Design Tech there from July 2026, with a visiting scholar stint at UC Berkeley EECS in 2026. His public talks include 'Lensing the Imagination' at Upper Bound 2025 and a Northwestern NICO seminar on tracing and shaping paths in design space, both given with a Midjourney affiliation.
What they run now
Academic now: AI-supported creativity, author intent in AI-assisted narrative, and evaluation methodology for creative systems. Recent work includes 'Can LLMs Generate Good Stories?' (CoG 2025), 'Bonsai: Cultivating Author Intent in LLM-Based Interactive Digital Narratives' (ICML 2026), and the homogenisation-analysis line showing LLMs narrow the range of human ideas. He retains a co-author relationship with Midjourney people through LiteraryTaste.
What they buy
Little, now. As an academic he specifies small research datasets and would be a grant-funded, low-value customer at best. His value to Julian is as a source and a credibility asset, not a buyer.
Career
- 2026-07 – presentAssistant Professor of Design Tech, Cornell Tech — Visiting Assistant Professor July 2025 – June 2026
- 2023 – 2025Research Scientist & Director, Storytelling Lab, Midjourney — Per his own CV; the index's 'Researcher, Midjourney' title is out of date
- 2026 – 2026Visiting Scholar, UC Berkeley, EECS
- 2022 – 2024Assistant Professor, Computer Science & Engineering, Santa Clara University
- 2022UC Santa Cruz, PhD, Computational Media
- 2020UC Santa Cruz, MS, Computational Media
- 2017University of Southern California, BA, Interactive Entertainment (minor: Game User Research)
On the record
Argues large language models narrow the range of ideas people generate, and proposes homogenisation analysis to measure it. source
Co-author on the finding that people diverge on creative preference and that stated preference poorly predicts revealed preference. source