DuckAI 2025 The 1st iCNS/ECE Symposium on AI Research and Innovations

Join this session to learn how to get started with Google's latest AI tools and models, including Gemini, Gemma, JAX, and Keras. We'll cover the latest models, and how to get started.

Short Bio: Josh Gordon is a Senior Staff Developer Advocate at Google, currently focusing on AI Frameworks (JAX, Keras, and PyTorch / XLA). He's also an adjunct professor at two schools - Columbia University, and Pace University - where he occasionally teaches graduate level courses on machine learning. He has over 15 years of machine learning experience to share.

Yufeng is a Developer Advocate focusing on machine learning libraries and frameworks. Yufeng has been at Google for eleven years, and in previous roles at Google, Yufeng built systems to create and curate bespoke datasets for LLMs, created the YouTube series "AI Adventures" at yt.be/AIAdventures, and has spoken at conferences around the globe, including Google I/O, Cloud Next, O'Reilly AI and Strata, PyCon, PyData, ML Prague, dotAI, and OSCON. He holds a Bachelor's degree in Biomedical Engineering from Johns Hopkins University. He is an avid marathoner and also plays the violin in the NYC Googler Orchestra.

Data Prep Kit ( https://github.com/data-prep-kit/data-prep-kit ) is built to support both structured and unstructured data across diverse domains, It enables development team to focus on model development by reducing the time and effort required for data preprocessing. Santosh will take your though the challenges in pre training data and how DPK is used to solve those challenges. Data Prep Kit is a open source project, Santosh will also go through the open source contribution opportunities for participants.

Short Bio: Santosh Borse is an experienced software engineer with 20 years of expertise in designing, developing, and managing complex software systems. He currently works as a Senior Research Engineer @ IBM Research, where he focuses on preparing and processing data for training the IBM Granite series models. Throughout his career, Santosh has held roles ranging from Junior Developer to Architect, contributing to impactful projects across AI, NLP, IoT, cloud computing, and more. He is also a named inventor on several granted patents in emerging technologies. Driven by a passion for problem-solving and innovation, Santosh believes in making the world a better place through software—a principle that continues to guide his work every day.

In modern large-scale networked cyber-physical systems such as smart grids, robotic networks, and supply chain networks, ensuring optimal performance while maintaining scalability and resilience is a critical challenge. Although powerful, classical optimal control techniques are often centralized and assume full system knowledge, making them ill-suited for decentralized implementation and online adaptation in dynamic environments. This talk presents a novel Graph Recurrent Neural Network (GRNN)-based framework for distributed optimal control in networked cyber-physical systems (N-CPS). Unlike prior approaches that either lack true decentralization or rely on offline training, our method enables online, fully distributed learning of optimal control laws using only local information and local communication among subsystems. We cast the problem as a self-supervised learning task and develop a consensus-inspired distributed training algorithm that allows each agent to independently adapt its controller weights over time. We establish local closed-loop stability guarantees to ensure safety and predictability by leveraging sector and slope restrictions on the nonlinear activation functions used in the GRNN. Numerical results have demonstrated the scalability and real-time training capabilities of the proposed method. The talk will also present ongoing efforts to generalize this framework to nonlinear settings and integrate it into a broader control-topology co-design paradigm for N-CPS.

Short Bio: Shirantha Welikala is currently an Assistant Professor in the Department of Electrical and Computer Engineering, Stevens Institute of Technology, Hoboken, NJ, USA (joined Fall 2023). He received the B.Sc. degree in Electrical and Electronic Engineering from the University of Peradeniya, Peradeniya, Sri Lanka, in 2015 and the M.Sc. and the Ph.D. degrees in Systems Engineering from Boston University, Brookline, MA, USA, in 2019 and 2021, respectively. From 2015 to 2017, he was a Temporary Instructor/Research Assistant in the Department of Electrical and Electronic Engineering at the University of Peradeniya, Sri Lanka. From 2021 to 2023, he was a Postdoctoral Research Fellow in the Department of Electrical Engineering, University of Notre Dame, Notre Dame, IN, USA. His main research interests include control and optimization of cooperative multi-agent systems, control of networked systems, passivity-based control, control and topology co-design, machine-learning, robotics, and smart grid. He has received several awards, including the 2015 Ceylon Electricity Board Gold Medal, the 2019 and 2023 President's Awards for Scientific Research in Sri Lanka, the 2021 Outstanding Ph.D. Dissertation Award in Systems Engineering at Boston University, and the 2022 Best Paper Award at the 30th Mediterranean Conference on Control and Automation held in Athens, Greece. For more information, please visit http://www.shiranthawelikala.com.

This talk highlights the role of data engineering in the fast-paced environment of hedge funds. It covers how engineers design and maintain scalable data pipelines, ensure data quality and integrity, and support a wide range of teams from research to compliance. Learn how modern data infrastructure enables better decision-making in the financial industry.

Short Bio: Zengye is a full stack quantitative engineer---frontend, backend, site reliability, distributed computing, model implementation. He previously worked at multiple hedge funds and is experienced with production systems.