Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
A developer detailed an eight-stage interview process at a fintech company, covering Python internals, rate limiting, and ML CI/CD. The process included a Demo Day, live coding, and a final round, with questions on TypeV…
An engineer argues that LLM evaluation is fundamentally a parameter sweep and should be treated as such, using tools like pydantic-evals and oryxflow instead of SaaS platforms. The post demonstrates a classification eval…
A new method from Amazon scientists, presented at the International Conference on Machine Learning (ICML), improves LLM-as-a-judge evaluation by accounting for correlations between judges' outputs, outperforming the best…
Google engineers reported a 4.7x faster prefill and 3.1x faster decode for Qwen 3.5-397B-A17B on Ironwood TPUs between April and June, achieved by using data parallelism for attention layers and expert parallelism for Mo…
A new AI memory system achieved #1 on the BABILong benchmark at 10M tokens, with a gaming audit showing 24/24 accuracy at 1M tokens and 16/16 at 2M, 5M, and 10M tokens on RULER variable tracking. The system, developed by…
Google introduced Gemini 3.5 Transcribe, its most precise speech-to-text model, available via the Gemini API in Google AI Studio and Gemini Enterprise Agent Platform. The model achieves a 4.0% word error rate for streami…
Google introduced Gemini 3.5 Transcribe, its most precise speech-to-text model, which achieves a 4.0% word error rate for streaming and 2.6% for non-streaming use-cases, and is now powering Gboard Rambler on Android, the…
Cigna Healthcare projects $200 million in medical cost reductions over the next three years from expanding its AI-enabled predictive tools across care management programs, with engaged customers saving an estimated $2,00…
Hugging Face contributor Blackroot published an explainer on August 26th detailing DeepSeek's Engram architecture, which uses hashed N-gram lookup tables to store recurring language patterns, reducing the load on mixture…
Amazon Web Services (AWS) announced that the Amazon SageMaker AI Python SDK v3 introduces a unified ModelTrainer class for training and ModelBuilder for deployment, replacing framework-specific estimators like SKLearn, P…
Microsoft researchers propose treating the harness boundary as an integration point for training modern AI agents, addressing the challenge that harnesses own tools, context, and control flow. The paper is featured as th…
Kathleen Creel, formerly assistant professor of philosophy and computer science at Northeastern University, has joined the University of Michigan as associate professor of philosophy, effective August 25. Her research ad…
Noteworthy AI, founded by CEO Christopher Ricciuti, raised $15,009,995 in an equity offering, according to an SEC Form D filed August 26, with 18 investors and a target of $18,234,987, leaving $3,224,992 available. The N…
Google has introduced Gemini Audio, a trio of speech models including Gemini 3.5 Transcribe, Gemini 3.5 Live, and Gemini 3.5 Live Experimental, designed to improve real-time dialogue and speech recognition. Gemini 3.5 Tr…
Amplify Investments' Amplify AI Powered Equity ETF (NYSEARCA:AIEQ), the first fully AI-driven ETF, has gained 100.2% since its Oct. 17, 2017 inception, while the S&P 500 gained roughly 199.7% over the same period, showin…
Google Cloud has integrated native TPU support into vLLM, the open-source LLM serving engine, to enable enterprise-grade precision for long-context multimodal embedding inference, targeting the Qwen3 Embedding model seri…
Caltech Professor Anima Anandkumar developed FourCastNet, the first open-source AI weather model, which is competitive with the best physics-based simulations and can run on consumer-grade GPUs. Anandkumar pioneered Neur…
Levels.fyi's compensation analysis for Lets Data Science found that the median AI pay premium over non-AI peers has stopped growing, with an L3 ML/AI engineer earning 1.34 times a non-AI peer in late 2024 and 1.29 times …
A developer shared a Python script for building a machine learning pipeline for real estate price prediction. The script includes exploratory data analysis, outlier filtering, feature engineering with KMeans clustering, …
OpenAI researchers developed scalable long-horizon reinforcement learning environments for browser use from real user traffic, filtering several hundred million task submissions down to about ten million distinct tasks a…