Speaker: Eric Slyman, Applied Research Scientist, Adobe Inc. 

Date and Time: Jan. 30, 2026

Zoom: https://oregonstate.zoom.us/j/94868713365 

Location: LINC 268

Title:
Trustworthy Creative AI: Practical Safety Mechanisms in Cross-Functional Systems

Abstract:

In industry, AI Safety rarely maps cleanly to one metric or vetting of some formal "golden" checkpoint. It is an evolving multi-stakeholder system including research, applied ethics, product, legal, and more. Done well, the major task of an AI Safety team is to navigate the often-conflicting prioritizations of these groups to discover an optimal compromise subject to the technical limitations of current practice and hardware. In the first part of this seminar, I will offer one perspective on how safety is operationalized in generative AI for creative workflows under such systems. We will walk through the development pipeline at a high level, discuss how these stakeholders interact, and describe concrete safety mechanisms that can be applied across the lifecycle, from data and training through product integration, deployment, and monitoring.

In the second part of the seminar, I will shift to cover my learnings as a recent graduate. Drawing on my own transition from academia-to-industry, I'll share advice on internship and job searches, how to translate academic experience into credibility signals across different roles, and “behind-the-scenes” insights from conducting technical interviews. Finally, I will argue that communication and other “soft skills” are often decisive in the job-search process and worth a deliberate investment.

Bio:

Eric Slyman is an Applied Research Scientist at Adobe, where he works on model evaluation, AI Safety, and intelligent prompting for Firefly’s multimodal generation and editing systems. Eric contributed to the newest generation of Firefly models, with an emphasis on reconciling core safety requirements with good-faith user intent in creative workflows. His cross-functional work includes regular collaboration with research, legal, and ethics teams to ship high quality models with low-intrusion safety mechanisms.

Previously, Eric earned his PhD in Artificial Intelligence and Computer Science from Oregon State University, where he studied fairness and reliability in large-scale vision-language systems under the practical constraints of ML engineers seeking to deploy their models. Eric’s research has been published at top-tier venues such as CVPR, ICCV, and NeurIPS, and has been shaped by industry relationships including Adobe, Apple, Google, and NVIDIA. He regularly speaks on responsible, reliable GenAI in practice, and his work has been featured by outlets such as Oregon Public Broadcasting and National Public Radio. In his time at OSU, Eric co-led OSU’s AI Graduate Student Association and helped found the OSU AI Application Support Program.