Articles from Source: Dropbox-Tech-Blog

At Dropbox, we recognize the challenge of maintaining visibility of security requirements during the development process. Our research revealed that only 12% of pull requests link back to their original threat models, causing potential gaps in security. To address this, we developed a system combining Model Context Protocol, foundational models, and Dash to retrieve relevant threat models during code reviews. This helps ensure security requirements are upheld throughout development. Learn...

🚀 Dropbox is evolving its approach to engineering productivity by transitioning from AI coding tools to agentic systems that execute scoped tasks. Traditionally, AI tools helped speed up coding, but they also created new bottlenecks in the development lifecycle. The focus is now on enhancing the entire software development process to handle increased output. Our internal platform, Nova, exemplifies this change, allowing engineers to work more efficiently while adapting to new workflows....

🚀 Introducing Nova, our new internal platform designed to enhance software development through coding agents. Nova enables engineers to run multiple coding sessions simultaneously and incorporates AI agents into automated workflows. This approach supports various engineering tasks, helping to reduce repetitive work and improve focus on critical projects. By using a platform solution, we can experiment with AI's role across the entire software development lifecycle, rather than relying on...

Magic Pocket, Dropbox's core storage system, has undergone improvements to enhance storage efficiency. By updating its compaction process and monitoring, the system can better adapt to workload changes. These changes address an increase in fragmentation and overhead caused by prior adjustments in data placement. The new multi-strategy approach aims to reduce storage overhead effectively. Learn how these advancements maintain durability and scalability at Dropbox. 📦🔧 #Dropbox #DataStorage...

At Dropbox, our server monorepo is central to product development, housing multiple services and libraries. However, as it grew to 87GB, cloning the repository took over an hour, slowing our engineering processes. To tackle this, we reduced the size by 77% to 20GB, cutting cloning time to under 15 minutes. This change aims to enhance developer velocity and streamline workflows. #DeveloperExperience #Monorepo #Dropbox #TechUpdates #SoftwareDevelopment 🚀💻📦

🚀 Exciting advancements in optimizing Dropbox Dash's relevance judge using DSPy! This open-source framework transformed the manual prompt engineering process into a measurable optimization loop. This led to improved task performance and reliability in production. The article discusses how DSPy helped address challenges like prompt brittleness and scaling relevance label generation effectively. For more insights, check out the full article! 📊🤖 #DropboxDash #DSPy #AIOptimization...

🔍 Dropbox Dash enhances search relevance using a blend of human and LLM-assisted labeling. When a user queries Dash, it retrieves relevant company information and uses it to provide tailored responses. The search model learns from human-annotated examples, ensuring high-quality results. By combining human insights with AI efficiency, Dash effectively improves the search experience for users. #Dropbox #AI #SearchRelevance #TechInnovation #MachineLearning

Advancements in AI are reshaping industries, with large models like Kimi-K2.5 and GLM-5 leading the way. These models excel in various applications but demand significant memory and power. ⚡ To address these challenges, low-bit inference techniques are crucial. They enhance efficiency by reducing the resources needed for AI tasks. Dropbox Dash exemplifies this, delivering quick AI-driven insights while managing resource use effectively. 📊 Explore the landscape of low-bit compute and its...

🚀 Dropbox is enhancing engineering productivity through AI tools like Claude Code and Cursor. At a recent executive roundtable, CTO Ali Dasdan and tech leaders discussed the effective integration of AI in workflows. Key insights included the importance of aligning AI adoption with business goals and the need for tailored solutions to fit Dropbox's unique challenges. Most developers are now utilizing AI tools, leading to increased code throughput. #AI #EngineeringProductivity #Dropbox...

🚀 Engineering VP Josh Clemm recently shared insights on Dropbox's use of knowledge graphs, MCP, and DSPy during a session in Jason Liu’s RAG course on Maven. He discussed how Dropbox Dash integrates with third-party apps, allowing users to easily search and manage their content. For a deeper understanding, check out the full video on Maven! #DropboxDash #KnowledgeGraphs #TechInnovation #MavenCourse #AI

🚀 Dropbox Dash is leveraging AI to enhance how we search and organize our work. By using a feature store, Dash ranks and retrieves relevant files quickly, adapting to user behavior across various content types. This system is essential for effective collaboration and improves access to important documents and conversations. The development of this feature store was tailored specifically for Dash, addressing unique infrastructure challenges and ensuring speed and efficiency. #DropboxDash #AI...

This summer, Dropbox welcomed 43 interns as part of the 2025 Camp Dropbox Intern Program. 🌍 They came from 27 colleges, including six international institutions. Over 12 weeks, interns worked on impactful projects, especially related to Dropbox Dash, our AI-powered search tool. They received over 6,000 hours of mentorship and participated in various community-building events. 🤝 Interns shared their experiences, highlighting technical achievements and personal growth throughout their time at...

🚀 Exciting advancements in AI are here with Dash! Initially a traditional search system, Dash has evolved into an agentic AI that interprets and acts on user requests. This shift emphasizes the importance of context engineering—delivering the right information at the right time. As Dash incorporates new tools, the focus remains on providing relevant context to enhance decision-making. With these developments, teams can streamline their workflows and improve project management. #AI...

🚀 Exciting news from Dropbox! They've partnered with Mobius Labs to enhance Dropbox Dash’s AI capabilities. With the integration of Aana models, Dash will improve its understanding of photos and videos, making it easier for teams to access and analyze their multimedia content. This multimodal processing will help users get answers faster without sifting through multiple tools. Stay organized and efficient with Dropbox Dash! #Dropbox #AI #MobiusLabs #Multimodal #Productivity

🚀 Explore Half-Quadratic Quantization (HQQ), a new method for compressing large AI models without needing calibration data. This technique significantly speeds up quantization, processing models like Llama-2-70B in under 5 minutes—over 50x faster than traditional methods. HQQ maintains competitive compression quality, enabling efficient deployment of large language models. #MachineLearning #AI #Quantization #HQQ #Llama2

Building effective conversational AI involves more than just training models. At Dropbox, we discovered that structured evaluation is crucial for accuracy. 🛠️ Initially, our evaluations were ad-hoc, but as we refined our processes, we developed a standardized approach to assess every change rigorously. This evaluation-first mindset is now part of our workflow. 📊 We created a playbook detailing datasets, metrics, and tools to help others implement this method across various media types,...

🚀 Hack Week 2025 at Dropbox focused on the theme “Keep It Simple,” encouraging innovation and collaboration. One highlight was the "Liquid Cooling CPU/GPU Hardware" project, which won the Learn Fast award. The team developed a custom liquid cooling system for GPU servers to address the thermal demands of AI workloads. This system demonstrated significant benefits, including lower operating temperatures and quieter performance. Future plans include expanding testing across multiple data...

🚀 AI is transforming how Dropbox operates, enhancing productivity for engineers. In early 2025, a shift occurred, leading to over 90% AI adoption among engineering teams. This change was driven by leadership alignment and a culture of knowledge-sharing. CTO Ali Dasdan emphasizes AI's role in automating tasks and allowing deeper focus on coding and innovation. #AI #Dropbox #Productivity #Innovation #TechTrends