TechPrufer
Sign inGet started

Blogs›Dropbox›AI Agent Platform for Engineering

AI Agent Platform for Engineering

AI Agent Platform for Engineering

3
posts
2026

Dropbox is evolving its engineering practices beyond AI code generation tools to AI agent platforms like Nova. Nova allows engineers to describe tasks in natural language, and AI agents execute these tasks within a controlled environment, interacting with the codebase, running tests, and iterating on failures. This platform aims to accelerate the broader software development lifecycle by handling scoped tasks, including feature development, migrations, test remediation, and bug investigation. The relevance judge for Dash has been optimized using DSPy, enabling a transition to a more cost-effective model while maintaining or improving accuracy by systematically adapting prompts and leveraging structured feedback for iterative refinement. This optimization process minimizes disagreement with human relevance judgments and ensures output reliability.

2026

Beyond code generation: rethinking engineering productivity in the age of AI agents

5/28/2026

This post introduces Dropbox's approach to AI agents as a platform for engineering productivity, moving beyond simple code generation. It details the development and impact of Nova, an internal AI agent platform that enables engineers to delegate scoped tasks to AI agents. The post highlights how Nova integrates with existing engineering workflows, handles tasks like code implementation, migrations, and bug investigation, and emphasizes the importance of surrounding systems (codebase context, internal practices, safe execution, workflow integration, human review) for agentic success. It also discusses the evolution of engineering productivity metrics to encompass workflow adoption, production output, and customer impact, rather than just code output, and stresses the need for enablement alongside tooling for successful agentic workflow adoption.

Introducing Nova, our internal platform for coding agents

5/21/2026

This post introduces Nova, Dropbox's internal platform for running coding agents. It details the platform's architecture, design choices (platform approach over single-purpose solutions), and its integration with Dropbox's development environment (monorepo, Bazel). It highlights key use cases like CI failure remediation, flaky test remediation (integrating with Athena), and migrations/dependency upgrades (integrating with RenovateBot). The post also discusses the importance of validation, context handling, and the evolution of agent capabilities beyond simple code generation.

How we optimized Dash's relevance judge with DSPy

3/17/2026

This post details the optimization of Dropbox Dash's relevance judge using the DSPy framework. The primary technical contribution is the application of DSPy's GEPA optimizer to adapt a relevance judging model (originally tuned for a proprietary model) to a more cost-effective open-weight model (gpt-oss-120b). This involved defining a clear objective based on human agreement (NMSE) and structural reliability, and using DSPy's reflection loop to generate structured feedback for iterative prompt refinement. The post also outlines the process of measuring human agreement and the challenges of prompt brittleness when migrating between models.