Ilkay Altintas

The Genesis Mission, launched by Executive Order, aims to accelerate AI-driven scientific discovery & innovation in the US. National labs, including DOE labs, will play a key role in providing HPC resources, research facilities & expertise to drive progress! #GenesisMission #AI #ScientificDiscovery #NationalLaboratory #HPC #DOE

9

"To ensure our use of #LLMs does not degrade science, we must use them as zero-shot translators: to convert accurate source material from one form to another." https://lnkd.in/gbT42f9A

1

The American Science Cloud marks an important step in building data foundations for artificial intelligence across the national laboratory system. I appreciate the collaboration with the U.S. Department of Energy (DOE) and the many teams working to make high-quality, AI-ready data safely and securely available to researchers.

119

Check out two pre-publications available in bioRxiv, by lead-author Jacob Croft, from Kelly Lee's lab, at the University of Washington in collaboration with our Center’s Computational Biology Core scientists, Hung Do and S. Gnanakaran from Gnana's lab at Los Alamos National Laboratory.   Croft et al. used cryo-electron tomography and sub-tomogram averaging of HIV-1 virus-like particles (VLPs), Hydrogen/Deuterium-Exchange Mass Spectrometry and Molecular Dynamics to capture the structural dynamics of HIV-1 Env proteins in the context of the cell membrane. They identified critical interactions and Env structural rearrangements providing insight to essential roles the membrane plays in presenting the Envs in its native conformation on the virion surface and dig deeper in a follow-up paper, revealing critical interactions inside the cell, between the Env cytoplasmic tail and matrix domain of Gag polyprotein, required for Env assembly, presentation, clustering, receptor binding and membrane fusion, all essential for viral propagation.   These exciting results reveal the power of these synergistic approaches and the importance of studying Env structures in the context of a membrane. Inside the cell, Gag polyproteins ,recruited to the inner leaflet, help package and present the Env on the outside cell surface. The cell now poised to infect new cells while, simultaneously, exposing the Env to an unrelenting barrage of antibodies seeking to neutralize further infection. Paper #1 https://lnkd.in/eizZwbJF Follow-up Paper https://lnkd.in/exxQxax9

13

This search-friendly guide explains the Berkeley Lab Staff Scientist role in nanofabrication—scope, eligibility, salary, work mode, and key dates—so applicants can prepare a strong dossier. It distills the official details into clear, professional language while keeping the structure intact and the focus on quantum materials and facility-user impact.

26

Interesting research co-authored by Green Imaging research collaborator Shaina Kelly, PhD and her colleagues from Columbia University and folks from Pacific Northwest National Lab. Great to see NMR being used in conjunction with other tech to look at these complex interactions related to CO2 storage. #NMR #carbonstorage https://lnkd.in/eUDn8qdr

10

1 Comment

“There's a potential to be actually able to design that material on a computer.” Dr. Sudip Dosanjh of Berkeley Lab's National Energy Research Scientific Computing Center (NERSC) talks AI, HPC, and the future of discovery in this 8-min interview. 🎧 Listen ⬇️

21

Lawrence Livermore National Laboratory will be using AI troubleshooting to improve operations at the National Ignition Facility (#NIF). It will be interesting to see how much this improves their operations as we move closer to sustainable #fusion reactions. Check out the article here: https://lnkd.in/eeuAayP2

19

The AAEES WORKSHOP on AI and ML for Environmental Applications picks up in the afternoon with a discussion on "Multivariate Polynomial Regression (MPR) for Response Surface Analysis (And Why Neural Networks Should Make You Nervous),” led David Vaccari, Stevens Institute of Technology.

3

The #DOECSGF builds community that fosters important scientific collaborations, such as the new Open Molecules 2025 dataset. Four alumni were critical to today's groundbreaking release: Berkeley Lab's Samuel Blau, Aditi Krishnapriyan (also University of California, Berkeley), and Santiago V.; and AI at Meta's Zachary Ulissi. https://lnkd.in/gnGuY5ui #computationalchemistry #DFT #machinelearning #HPC Harvard University, Stanford University, UCLA, Massachusetts Institute of Technology, Berkeley Lab Computing Sciences

105

2 Comments

Tipping as Fiber Degeneration: A Grothendieck-Theoretic Foundation of Admissibility Loss Zenodo DOI: https://lnkd.in/ebPu5ka7 This manuscript develops a categorical and geometric reformulation of tipping points based on admissible fibrations over forcing histories. Rather than organizing tipping by mechanism (bifurcation-induced, rate-induced, stochastic, or interaction-driven), collapse is defined as boundary contact in admissible state space. Forcing histories form a base category; admissible states form fibers over that category; Distance to Admissibility (DtA) provides a canonical boundary metric; Joint Margin encodes interaction curvature under multi-source forcing. Within this structure, classical tipping mechanisms appear as projections (base-change views) of a single underlying admissibility geometry. Using the Atlantic Meridional Overturning Circulation (AMOC) as a worked example, the paper demonstrates how bifurcation, rate effects, noise-induced collapse, and multi-driver interactions reduce to degeneration of admissible fibers under forcing histories. The framework is not climate-specific. Any forced dynamical system with a collapse criterion admits the same geometric organization. I believe tipping theory benefits from shifting from mechanism-centered classification to boundary-centered geometry. Happy to hear thoughts, critiques, and discussions. Louis Nguyen, 2026

1

Problems that stalled quantum computers for years? Argonne cracked them by letting neural networks learn differential equations-and the math community noticed. At the AN25 conference in Montreal, Argonne National Lab researchers dropped some serious breakthroughs. Emil Constantinescu showed how neural ordinary differential equations can slash computational costs. Same accuracy. Way less processing power. But that's not all: • Jeff Larson tackled quantum error correction with mixed-integer models • Matt Menickelly pushed quantum sensor precision limits • Vishwas Rao scaled signal processing to hundreds of millions of variables That last one hits GPS, medical imaging, and X-ray crystallography. Big impact. The GPU-accelerated methods are game-changers. We're talking real-world applications that were bottlenecked for years. What excites me most? These aren't lab curiosities. They solve actual problems. Quantum computing needs error correction to be viable. Signal processing powers everything from your phone's GPS to cancer imaging. Argonne just moved the needle on both. The optimization community is paying attention. And they should be. Which breakthrough do you think will have the biggest real-world impact? #QuantumComputing #Optimization #TechInnovation 𝗦𝗼𝘂𝗿𝗰𝗲꞉ https://lnkd.in/esPeCbJP

1

1 Comment

We are pleased to release NeurDB v0.5.0, bringing substantial enhancements in AI-powered data analytics and autonomous operations. 🔑 Key Features ✅ In-Database Model Management Integrated a database-native deep learning model store (NeurStore) for efficient model persistence and retrieval. Optimized model loading and saving mechanisms with reduced storage overhead and improved I/O performance. Enhanced metadata tracking to support multi-version model management and dependency tracing. ✅ Concurrency Control Enhancement Refined the NeurDB Access Method for improved transaction throughput under mixed workloads. Enhanced lock management and state abstraction for learned concurrency control. Added test cases to validate transaction isolation and detect potential anomalies. ✅ KV-based Storage Engine Enhancement Integrated an inter-process communication (IPC) channel to reduce context-switch overhead and improve latency. Added test cases to validate correctness under concurrent read/write. ✅ Automatic Data Preparation Pipeline Introduced an automated data preparation and feature transformation pipeline for AI analytical queries. Supports training-ready dataset generation. https://lnkd.in/gc_9RmTA #AI #Data #NeurDB #ZhejiangUniversity #Ningbo

18

A new feature from Nature titled "Quantum computers will finally be useful: what’s behind the revolution" highlights research from Princeton ECE professors Andrew Houck and Nathalie de Leon. The piece describes how their work improving the performance of qubits has contributed to dramatically shortening the projected timeline for practical quantum computers. "Just a few years ago, many researchers in quantum computing thought it would take several decades to develop machines that could solve complex tasks, such as predicting how chemicals react or cracking encrypted text," according to the article. "But now, there is growing hope that such machines could arrive in the next ten years." Check it out 👉 https://lnkd.in/eC5cSDs4

19

1 Comment

Excited to TA [these courses](https://lnkd.in/eja-bBiy) over IAP led by Shaohao Chen of the [MIT Office of Research Computing and Data](https://orcd.mit.edu/). On Day 2 (Thursday, January 15) of Intro to Parallel Programming, we will cover: - GPU basics - CUDA programming - Deep learning on a GPU - Distributed deep learning

13