2026_04_15.html

---
layout: talk
title: "The Agentic Revolution: Lightning Sessions on Agentic Workflows"
authors: Woong Shin, Jan Janssen, Du Ming, Xiangyu Yin
event_date: April 15, 2026
times: 11:00am PST / 2:00pm EST / 20:00 CEST
talk_number: 13
speakers: [woong_shin, jan_janssen,ming_du, xiangyu_yin]
given: false
image: /images/talks/agentic_banner.jpg
<!-- presentation: /files/talks/20260415-AgenticRevolution.pdf -->
<!-- video: https://www.youtube.com/embed/TPAz4bbmW_0?si=JO3ARdoIYRji-TKl -->
---

<strong>The (R)evolution of Scientific Workflows in the Agentic AI era: Towards Autonomous Science</strong><br />
<em>Woong Shin (Oak Ridge National Laboratory)</em>

<br /><br />

Modern scientific discovery increasingly requires coordinating distributed
facilities and heterogeneous resources, forcing researchers to act as manual
workflow coordinators rather than scientists. Advances in AI leading to AI agents
show exciting new opportunities that can accelerate scientific discovery by
providing intelligence as a component in the ecosystem. However, it is unclear
how this new capability would materialize and integrate in the real world. To
address this, we propose a conceptual framework where workflows evolve along two
dimensions, intelligence (from static to intelligent) and composition (from
single to swarm), to chart an evolutionary path from current workflow management
systems to fully autonomous, distributed scientific laboratories. By embedding
reasoning and adaptation into workflows, these labs have the potential to
accelerate discovery by factors of 10 to 100, transforming exploratory science
into a continuous, machine-augmented process.

<br /><br />

<strong>Title: TBD</strong><br />
<em>Jan Janssen (Max Planck Institute for Sustainable Materials)</em>


<br /><br />

<strong>Agentic AI for experiments and data analyses at the APS</strong><br />
<em>Du Ming, Xiangyu Yin (Argone National Laboratory)</em>

<br /><br />

We will introduce the current efforts of using vision language model (VLM) agents
for automated and low-barrier beamline operations and data processing algorithm
research. We first present Experiment Automation Agents (EAA), an agent capable
of controlling beamline instruments and making decisions based on image semantics,
with a few cases demonstrating how it automates and democratizes experimental
operations at APS beamlines. We will then introduce works on agentic data
processing, which includes PEAR, a domain-expert system that tunes
ptychographic reconstruction hyperparameters using reconstructed image as
feedback, and Pty-Chi-Evolve, an auto-research agent that autonomously searches
for regularization operators during iterative reconstructions to enhance result
quality. 

<br /><br />