engblogs

Insights on LLM Agents, Evaluation, and Future Architectures from an Interview with Graham Neubig

Reduce LLM parameter counts and memory usage with LoRA: Low-Rank Adaptation for fast customization and fine-tuning of large language models

Empowering product teams with three key principles for making impactful decisions and building user-centric products at Duolingo

Efficient low-rank adaptation technique, LoRA, reduces complexity of fine-tuning large language models by approximating updates with low-rank matrices for significant memory and time savings.

Enhancing data reporting for New York Times Games with revamped data architecture and dashboard suite

Efficiently fine-tune large language models using Low-Rank Adaptation (LoRA) to reduce trainable parameters by 10,000 times and GPU memory requirements by 3 times.

Google Cloud introduces Gemma, a family of lightweight open models for AI development on Vertex AI and GKE.

Enhance runtime security monitoring for Google Cloud services with Falco and gcpaudit plugin

Enhance user experience with new Autocomplete Preview featuring Address Validation, intuitive pricing, and expanded place types from the Places API

Orchestrate gen AI models with parallel steps using Workflows for efficient processing.

Optimizing digital exchanges for performance, scale, and resilience on Google Cloud

Exploring the latest advancements in private training for on-device language models, focusing on differential privacy guarantees and federated learning in Gboard production.