Sangbae Kim

A team led by researchers in UC Berkeley’s Department of Materials Science and Engineering (led by Prof. Rayne Zheng and postdoc Marco Maurizi) has developed a new AI-driven method for the fully automated design of engineering materials with complex physical behaviors. The approach enables the inverse design of metamaterials for advanced functionalities such as energy absorption and sound cancellation. Notably, the AI model also accounts for processing-related defects, offering more robust and manufacturable solutions. This work is also featured on the front cover of Nature Machine Intelligence. Learn more in the Berkeley Engineering News article https://lnkd.in/e5DeuAat and our LinkedIn post https://lnkd.in/gpSApgeU

125

1 Comment

🚀 Excited to share Compressive Light-Field Tokens (CLiFT) — a new scene representation for neural rendering developed by my students (Zhengqing Wang, Yuefan Wu, Jiacheng Chen, Fuyang Zhang). CLiFT is a Light-Field Token, compressed for compute-efficient and adaptive neural rendering, achieving (2x to 16x) reduction in storage and (66% to 100%) increase in FPS, while preserving high visual quality with only (0.54dB to 3dB) drop in PSNR. 🔗 Project: https://lnkd.in/gBqvMWi8 📦 Code coming 8/1!

166

3 Comments

In this open-access article published today in npj 2D Materials and Applications, we report a high-throughput device simulation framework to study memristive and synaptic devices based on 2D materials. Thanks to Benjamin Spetzler and Martin Ziegler for this amazing collaboration! This comprehensive paper simulates thousands of use cases of a finite-element continuum model directly connecting fundamental materials properties such as vacancy mobility with a wide range of learning behaviors of 2D synaptic devices such as long-term potentiation and depression. Thus, this model of correlated electron and vacancy transport can simulate but goes beyond the standard dc memristive characteristics. Such an approach is needed because the temporally complex nature of synapses cannot be directly discerned by any other device measurements. This way diverse characteristics such as non-linearity in weight update protocols, drift, overshoot, etc. can be quantitatively correlated with materials parameters and may be useful in developing better 2D memristors and memtransistors. https://lnkd.in/gnaMhj5G

95

1 Comment

🚍 What if shifting a bus departure by just a few minutes could save an entire vehicle? That’s the idea behind our new paper, recently published in the INFORMS Journal on Computing. Together with Thomas van der Schaft, we developed a Dynamic Discretization Discovery (DDD) algorithm that tackles the Multi-Depot Vehicle Scheduling Problem with Trip Shifting (MDVSP-TS). Why this matters: - Even tiny timetable adjustments can unlock huge efficiency gains in public transport. - Our algorithm solves real-world bus networks with nearly 4,000 trips to optimality. - It outperforms existing exact methods by a wide margin. This work shows how smart optimization can make public transport more efficient, sustainable, and practical at scale. 👉 Read more here: https://lnkd.in/eS_dhZZ3

93

9 Comments

The Spring issue of IEEE Solid-State Circuits Magazine is out! https://lnkd.in/gxbjNtKw This issue is dedicated to Integrated Circuits for AI and ML applications, and was guest edited by Marian Verhelst and Charlotte Frenkel. It features four excellent articles that highlight core challenges and methodologies in AI hardware acceleration (see https://lnkd.in/gayteYmV for more detail). The feature articles are: 1) "Design and Optimization of Efficient Digital Machine Learning Accelerators" by Wei Tang, Sung-Gun Cho, Jie-Fang Zhang, and Zhengya Zhang examines AI acceleration through optimized computational kernels, such as matrix–matrix multiplication, leveraging architectures like SIMD and systolic arrays (10.1109/MSSC.2025.3549361). 2) "SRAM- and eDRAM-Based Compute-in-Memory Designs, Accelerators, and Evaluation Frameworks," by Yifan He, Xiaofeng Hu, Hongyang Jia, and Jae-sun Seo addresses in-memory computing (IMC) as a means to reduce data movement overhead, covering both analog and digital IMC approaches (https://lnkd.in/gHiukWKx). 3) "Open Source Heterogeneous SoCs for Artificial Intelligence," by Francesco Conti, Angelo Garofalo, Davide Rossi, Giuseppe Tagliavini, and Luca Benini explores the integration of heterogeneous systems on chip, demonstrating how open source platforms such as PULP support scalable AI hardware development (https://lnkd.in/gAQgK-8d). 4) " Sparsity-Aware Hardware," by Man Shi, Adrian Kneip, Nicolas Chauvaux, Jiacong Sun, Charlotte Frenkel, and Marian Verhelst discusses sparsity-aware hardware design, illustrating how structured and unstructured sparsity can be leveraged for significant performance improvements using an adapted roofline model (https://lnkd.in/gv8eTyvd). The issue also features regular columns: 1. Behzad Razavi’s “The Analog Mind” column (https://lnkd.in/gxKyqZr7) continues the series on CMOS inverter applications 2. Chris Mangelsdorf’s “Shop Talk: What You Didn’t Learn in School” column picks up where the previous issue left off, continuing the discussion on Miller compensation in "Miller’s Wrong Half-Plane Zero" (https://lnkd.in/gngTrywV). This time, the focus is on the right half-plane zero and its potential to degrade stability if not properly addressed. 3. Farhana S.’s “Circuits From a Systems Perspective” traces the historical evolution of wireless charging in everyday devices (https://lnkd.in/gnVhqUyF). Lot's of interesting articles to read! Enjoy! #ieeesscc #analog #ieee #ai #ml #solidstatecircuits #hardwareacceleration #semiconductors #inmemorycomputing

258

3 Comments

Expanding the frontiers of polymer design with physics-informed graph neural networks I’ve just published a new Substack post discussing recent work by Jiang and Webb, who integrate polymer physics with graph neural networks to predict properties across a wide range of architectures—including dendrimers—and chemical compositions. A key contribution of their physics-informed approach is its ability to extrapolate beyond typical training sets, paving the way for more efficient and accurate polymer design. Link to the post: https://lnkd.in/dCaXCewJ #PolymerScience #PhysicsInformed #GraphNeuralNetworks #MaterialDesign #DataScience #MachineLearning #Chemistry #AcademicResearch #PolymerInformatics #DeepLearning #NeuralNetworks #ComputationalChemistry #AIResearch #Dendrimer #ScienceInnovation

70

1 Comment

New Publication Alert ‼️ In our new paper published in the Journal Computer Methods in Applied Mechanics and Engineering, we propose a new computational framework that employs variational methods, specifically by leveraging the principle of minimum pressure gradient, to turn fluid mechanics problems into a minimization problem whose solution can be used to predict the flow field in unsteady incompressible viscous flows. This method exhibits two particularly intriguing properties: 1) It circumvents the chronic issues of pressure–velocity coupling in incompressible flows, which often dominates the computational cost in computational fluid dynamics (CFD). 2) It eliminates the reliance on unphysical assumptions at the outflow boundary, addressing another longstanding challenge in CFD. Kudos to Hussam Alhussein (sababha) for leading this transformative work, which resulted from a collaboration with the group of Prof. Haithem Taha at UC Irvine. #nyuad #nyu #fluidmechanics #ucirvine #fluids #CFD https://lnkd.in/dEGvTQxu

95

2 Comments

Have you wondered how to connect across different set of geometric design parameters that define specific architected materials? In this study published in Extreme Mechanics Letters, we introduce geometric descriptors--that unify various architectures while capturing underlying nanoscale size effects and nanofiber interactions in vertically aligned carbon nanotubes--in an artificial neural network framework to relate the structure to bulk properties. We test the model to predict properties of architectures that are even outside the training data (non-intuitive architectures) and demonstrate great agreement with experiments. Finally, we present a general framework on how these geometric descriptors can be used to study different architected hierarchical materials with underlying nano and mesoscale mechanisms, unifying their design space. Congratulations to PhD student, Bhanugoban Maheswaran, for leading this effort with our former team members Abhishek Gupta and Komal Chawla. Read the paper here: https://lnkd.in/g3H4By2k

67

We published a new paper in the Journal of Fluid Mechanics regarding the interpretation of sensor signals in high-speed boundary layers. We compute the domain of dependence (DOD) for wall-pressure sensors in a Mach 4.5 boundary layer using adjoint-sensitivity analysis. We analyze how sensor placement in laminar, transitional, and turbulent zones affects what flow features are actually "felt." Check out the full study in JFM: https://lnkd.in/gbii3pc9 In addition, thanks David Buchta for all the insightful discussions and sharing the computational expertise, and Curtis Haas for careful reading of the paper!

62

3 Comments

The final paper from Rachel Lim's time at Penn State was just published https://lnkd.in/egHb59wy. This work grew from nearly a decade of collaboration with Thien Phan. At the time, we were interested in using X-ray diffraction as a thermomechanical diagnostic for additive manufacturing, but it wasn't clear how we would ever be able to separate the effects of thermal and mechanical loading in the measured signal. With the help of Zi-Kui Liu and Shun-Li Shang, we trained a probabilistic machine learning model to deconvolute these two effects within experimental diffraction data (measured with Andrew Chuang at the Advanced Photon Source). While machine learning may not solve every problem, (when trained properly) it can be a valuable tool for solving complex inverse problems.

65

3 Comments