Quantization-aware training (QAT) is a leading technique for improving the accuracy of quantized neural networks. Previ- ous work has shown that decomposing training into a full-precision (FP) phase followed by a QAT phase yields superior accuracy compared to QAT alone. However, the optimal allocation of compute between the FP and QAT phases remains unclear. We …
Fraud continues to cause significant financial damage globally, with U.S. consumers alone losing $12.5 billion in 2024—a 25% increase from the previous year according to the Federal Trade Commission. This surge stems not from more frequent attacks, but from fraudsters’ increasing sophistication. As fraudulent activities become more complex and interconnected, conventional machine learning approaches fall short …
Read more “Modernize fraud prevention: GraphStorm v0.5 for real-time inference”
For AI agents to be really useful, they need to be able to securely interact with enterprise data. In July, we introduced a toolset to help AI agents interact with and analyze business data in BigQuery through natural language, and with just a few lines of code. Today, we’re taking the next step, with “Ask …
Read more “Forecasts and data insights come to BigQuery’s MCP and Agent Development Kit tools”
While federated learning (FL) and differential privacy (DP) have been extensively studied, their application to automatic speech recognition (ASR) remains largely unexplored due to the challenges in training large transformer models. Specifically, large models further exacerbate issues in FL as they are particularly susceptible to gradient heterogeneity across layers, unlike the relatively uniform gradient behavior …
By Jun He, Yingyi Zhang, Ely Spears TL;DR We recently upgraded the Maestro engine to go beyond scalability and improved its performance by 100X! The overall overhead is reduced from seconds to milliseconds. We have updated the Maestro open source project with this improvement! Please visit the Maestro GitHub repository to get started. If you find …
Read more “100X Faster: How We Supercharged Netflix Maestro’s Workflow Engine”
Today, we’re announcing the general availability of Claude Sonnet 4.5, Anthropic’s most intelligent model and its best-performing model for complex agents, coding, and computer use, on Vertex AI. Claude Sonnet 4.5 is built to work independently for hours, maintaining clarity while orchestrating tools and coordinating multiple agents to solve complex problems. It’s designed to excel …
Large foundation models are typically trained on data from multiple domains, with the data mixture—the proportion of each domain used—playing a critical role in model performance. The standard approach to selecting this mixture relies on trial and error, which becomes impractical for large-scale pretraining. We propose a systematic method to determine the optimal data mixture …
By Prudhviraj Karumanchi, Samuel Fu, Sriram Rangarajan, Vidhya Arvind, Yun Wang, John Lu Introduction Netflix operates at a massive scale, serving hundreds of millions of users with diverse content and features. Behind the scenes, ensuring data consistency, reliability, and efficient operations across various services presents a continuous challenge. At the heart of many critical functions lies …
Read more “Building a Resilient Data Platform with Write-Ahead Log at Netflix”
This blog was co-authored with Kuldeep Singh, Head of AI Platform at Innovaccer. The integration of agentic AI is ushering in a transformative era in health care, marking a significant departure from traditional AI systems. Agentic AI demonstrates autonomous decision-making capabilities and adaptive learning in complex medical environments, enabling it to monitor patient progress, coordinate …
Read more “Building health care agents using Amazon Bedrock AgentCore”
New era of physical agents will help robots perceive, plan, think, use tools and act to solve complex tasks.