engblogs

Understanding the basics of generative AI technology and its significance

A blogpost introducing Sliced Anti-symmetric Decomposition (SAD), a new model for collaborative filtering based on implicit feedback.

OpenAI's recent acquisition of Global Illumination brings new advancements to the field.

This blogpost discusses the implementation of FineRecon, a depth-aware feed-forward network that improves upon the coarse 3D geometry reconstruction typically generated by deep neural networks.

Introducing DNIKit, an open source Python framework for analyzing machine learning models and datasets, enabling comprehensive dataset analysis, finding near duplicate samples, uncovering rare data samples, annotation errors, or model biases, and compressing networks

Leveraging Selective Execution in Amazon SageMaker Pipelines to enhance efficiency in MLOps and streamline the productionization of machine learning models.

Three lessons for maximizing impact in innovation through experimentation at Spotify.

Enabling personalization without data replication through ActionIQ integration with Databricks Lakehouse.

Improving connection overhead and reducing churn in Zuul by implementing HTTP/2 multiplexing and partitioning the origin into subsets

The blogpost discusses the goal of enabling HTTPS-First Mode by default, including automatic upgrades to HTTPS, warning on insecure downloads, and expanding HTTPS-First Mode protections for different user groups.

A blogpost discussing the challenges of using data effectively at Instacart and the engineering initiative called 'Ground Truth' that aims to address these challenges by establishing robust data foundations and tools, with a focus on the next era of data at Instacart and the use of AI and LLMs to provide a richer self-serve experience with data.

Train self-supervised vision transformers on overhead imagery using Amazon SageMaker for aerial imagery model use cases.