10 posts · searchable
10 posts
Improves storage efficiency in Magic Pocket by turning compaction into a layered, adaptive pipeline with enhanced monitoring.
Reduces monorepo size to improve developer velocity without increasing friction.
Optimizes Dash's relevance judge using DSPy to automate prompt engineering and improve performance and cost.
Uses LLMs to amplify human labeling for training Dash's search ranking models.
Explains how low-bit inference enables efficient AI for products like Dropbox Dash.
Shares insights from an executive roundtable on AI and engineering productivity, including adoption of AI coding tools.
Engineering VP discusses how knowledge graphs, MCP, and DSPy are used in Dash.
Describes the feature store powering real-time AI for ranking and retrieval in Dropbox Dash.
Highlights the 2025 summer intern class and Dropbox's intern program focused on growth and innovation.
Explains how Dash uses context engineering to help AI models focus on relevant information.