TODAY: Guest Seminars at YSU Krisp AI Lab
14 views
Skip to first unread message
Tatev Vardanyan
Sep 19, 2025, 3:16:51 AMSep 19
to Machine Learning Reading Group Yerevan
Dear friends,
Seminar Abstract: Hospitals are complex, highly interconnected systems where dynamic, nonlinear patient flow can create persistent bottlenecks that often originate in non-priority wards and emerge in critical care areas. The nonlinearity, evolving nature, and unpredictable changes make it challenging to perceive system behaviour. To address this, I will present two network-agnostic computational approaches capable of modelling diverse hospital networks, regardless of topological changes. The first is a hybrid modelling approach that integrates agent-based network simulation with network algorithms to identify persistent bottlenecks arising from both ward-to-ward connectivity and patient flow, supporting evidence-based decision-making and performance evaluation under evolving hospital topologies. The second is a dynamic nonlinear flow algorithm, derived from the Ford–Fulkerson and Edmonds–Karp algorithms, modified for complex hospital systems to capture bottleneck persistency, severity, and root causes, as well as to achieve better computational efficiency. Together, these approaches offer flexible, scalable, and adaptable tools for analysing and improving patient flow in complex hospital systems.
-------------------
Presenter: Dr. Michel Gokan Khan, KTH Royal Institute of Technology (Google Scholar, Profile, Bio)
Seminar Title: AI-Driven Digital Twins in Action: From Cloud Native Systems to Advanced Industrial Applications
Seminar Abstract: Digital twins, while widely applicable, vary significantly in design and purpose depending on the context. In this talk, I will present recent work on how AI can automate their creation and how these distinct applications employ specialized technologies for unique goals. I will discuss our work in industrial digital twinning powered by sensor data, computer vision, 3D reconstruction, object detection, and 3D Gaussian Splatting, showing how they provide immersive visualization and actionable insights for operators and quality assurance teams, capturing spatial and procedural knowledge to enhance performance and skill transfer. I will also present our work on digital twins of cloud native systems, highlighting how they optimize microservice deployments, improve resource allocation, and enable seamless scalability. Drawing on both research and industrial experience, I will share projects, challenges, and future directions to illustrate how AI-driven digital twins can enhance operational agility and efficiency across domains.
We are hosting two seminars today at 4:30 PM featuring guest speakers from Stockholm University and KTH Royal Institute of Technology. Please see details about the presenters and topics below.
Presenter: Dr. Arsineh Boodaghian Asl, Senior Lecturer, Stockholm University Google Scholar, Homepage, Bio)
Seminar Title: Network-agnostic computational approaches to capture the dynamic nonlinear behaviour of complex hospital systems.
Seminar Abstract: Hospitals are complex, highly interconnected systems where dynamic, nonlinear patient flow can create persistent bottlenecks that often originate in non-priority wards and emerge in critical care areas. The nonlinearity, evolving nature, and unpredictable changes make it challenging to perceive system behaviour. To address this, I will present two network-agnostic computational approaches capable of modelling diverse hospital networks, regardless of topological changes. The first is a hybrid modelling approach that integrates agent-based network simulation with network algorithms to identify persistent bottlenecks arising from both ward-to-ward connectivity and patient flow, supporting evidence-based decision-making and performance evaluation under evolving hospital topologies. The second is a dynamic nonlinear flow algorithm, derived from the Ford–Fulkerson and Edmonds–Karp algorithms, modified for complex hospital systems to capture bottleneck persistency, severity, and root causes, as well as to achieve better computational efficiency. Together, these approaches offer flexible, scalable, and adaptable tools for analysing and improving patient flow in complex hospital systems.
-------------------
Presenter: Dr. Michel Gokan Khan, KTH Royal Institute of Technology (Google Scholar, Profile, Bio)
Seminar Title: AI-Driven Digital Twins in Action: From Cloud Native Systems to Advanced Industrial Applications
Seminar Abstract: Digital twins, while widely applicable, vary significantly in design and purpose depending on the context. In this talk, I will present recent work on how AI can automate their creation and how these distinct applications employ specialized technologies for unique goals. I will discuss our work in industrial digital twinning powered by sensor data, computer vision, 3D reconstruction, object detection, and 3D Gaussian Splatting, showing how they provide immersive visualization and actionable insights for operators and quality assurance teams, capturing spatial and procedural knowledge to enhance performance and skill transfer. I will also present our work on digital twins of cloud native systems, highlighting how they optimize microservice deployments, improve resource allocation, and enable seamless scalability. Drawing on both research and industrial experience, I will share projects, challenges, and future directions to illustrate how AI-driven digital twins can enhance operational agility and efficiency across domains.
Reply all
Reply to author
Forward
0 new messages