engblogs

Host and scale your applications close to your users with ease using Fly.io

Explore the importance of data normalization strategies for building a Cybersecurity Lakehouse and learn how to overcome challenges in this final part of the 'Cybersecurity Lakehouses Best Practices' blog series.

Challenges and solutions for utilizing enterprise-grade large language models (LLMs) in AI applications

Addressing enterprise challenges with large language models (LLMs) in AI through data-centric development, with a focus on customization and accuracy.

Learn how to optimize the scaling of your EMA model for improved training dynamics and performance in machine learning.

Automating Behavioral Testing in Machine Translation: Proposing a practical evaluation framework using Large Language Models (LLMs) to generate source sentences and verify behavior of MT systems.

Exploiting activation sparsity in large language models to reduce computation and weight transfer, using ReLU activations with minimal performance trade-offs.

Diffusion Models as Masked Audio-Video Learners: Exploring the synergy between diffusion models and MAViL for enhanced efficiency without compromising performance.

A blogpost proposing a novel architecture for flexible keyword spotting based on homogeneous audio-text embedding.

Advancing responsible AI through adversarial testing to mitigate safety concerns in generative AI models.

Overcoming enterprise challenges with large language models (LLMs) in AI and data-centric development.

Predicting the changes and evolution of French language in the next century, from negation to grammar rules and pronouns.