engblogs

Explore six deployment archetypes to meet reliability, cost, operational, and latency needs in Google Cloud.

Enhancing AI workload security with Confidential Accelerators for secure ML training and collaboration

Enhancing Google Cloud networking for generative AI applications with new capabilities and optimizations

Accelerate deep learning training and simplify orchestration with AWS Trainium and AWS Batch by seamlessly integrating powerful ML acceleration capabilities with efficient orchestration functionalities.

How Twilio leveraged Amazon SageMaker MLOps pipelines with PrestoDB for frequent model retraining and optimized batch transform

Utilizing advanced video-to-audio technology for synchronized soundtracks with control over audio output.

Optimizing drug discovery with cost-efficient synthesis plans using SPARROW algorithm

Enhancing automated speech recognition for virtual assistants through server-side rescoring of spoken entity-centric knowledge queries.

Utilizing hypernetworks for personalized adaptation of automatic speech recognition to atypical speech

Utilizing a Transformer-based model for ASR N-Best rescoring and rewriting to improve word error rate

Efficient fine-tuning techniques improve time-sensitive knowledge editing in Large Language Models.

Investigating weak-to-strong generalization in open-source models yields insights on improving generalization through log-confidence auxiliary loss.