Biomolecular structure prediction and co-folding with models like OpenFold3 are now mainstream, large-scale workloads powering drug discovery and protein design. Increasingly, they’re driven end-to-end by AI agents. For an agent to run that pipeline well, every step needs to be fast and scalable: Multiple Sequence Alignment (MSA) generation, co-folding inference, serving, and multi-GPU scale-out. A…
Topics Leadership Leadership Skills Column Our expert columnists offer opinion and analysis on important issues facing modern businesses and managers. More in this series Summary: MIT Sloan Management Review’s summer 2026 issue includes articles that point to the value of contributing to collective learning among industry peers. Articles on topics as varied as emerging technology,…
In object-oriented programming (OOP), an interface describes the methods a class must implement to play a specific role. This set of methods expresses the interface contract. Python has no dedicated syntax for interfaces, but it gives you two mechanisms for modeling them: abstract base classes and protocols. Python’s everyday duck typing takes advantage of both….
At the edge, our E2B and E4B models redefine on-device utility, prioritizing multimodal capabilities, low-latency processing and seamless ecosystem integration over raw parameter count. Powerful, accessible, open To power the next generation of pioneering research and products, we’ve sized the Gemma 4 models specifically to run and fine-tune efficiently on hardware — from billions of…
An artificial intelligence (AI) agent can process an invoice, help adjudicate a claim, or classify a support ticket in a proof of concept. But running these agents across thousands or even millions of work items in a production environment introduces an entirely different set of challenges. At enterprise scale, success depends on much more than…
The race to build AI factories is well underway, and the winning organizations have learned that AI data infrastructure determines whether or not GPU investments pay off, while others are still scrambling to assemble workable solutions. That divide is the clearest indicator from the field, said Alex Bouzari (pictured), chairman, co-founder and chief executive officer of…
When you’re building a computer vision system, the end goal isn’t the model but the problem it solves. That might be inspecting critical infrastructure to schedule maintenance crews more effectively, catching defects on a production line before they ship, or analyzing satellite imagery to evaluate crop health across a region. Solving those problems isn’t a…
The rise of the vector database As a result of the rapid advancement of generative AI in recent years, many companies are rushing to integrate AI into their businesses. One of the most common ways of doing this is to build AI systems that answer questions concerning information that can be found within a database…
TL;DR: DeepSeek-V4 support was live in SGLang on Day-0, but the Day-0 stack was only the starting point. Since launch, we have coordinated a set of kernel, runtime, and hardening improvements: MHC fusion and token-bucket prewarm, KV Compression V2, W4A4 MegaMoE, stronger SWA budgeting and eviction behavior, better disaggregated decode admission, breakable CUDA graph support…
Andrew covers the biopharma industry, scientific research, and public health across the continent. You can reach Andrew confidentially on Signal at drewqjoseph.71. This is the web edition of STAT’s AAIC in 30 newsletter. Sign up to get it in your inbox here. Welcome to the first edition of our pop-up newsletter from AAIC. I’m STAT’s…