Cohort profile and program design

Dropbox's Emerging Talent team brought 43 interns into the 2025 Camp Dropbox Intern Program, drawn from 27 colleges and universities, six of them international institutions in Canada, Poland and Ireland. Twenty-eight of the interns were placed on Engineering teams. The program ran 12 weeks, from May through September, under Dropbox's Virtual First culture.

Support included more than 6,000 hours of one-on-one mentorship. Interns worked on projects aligned with team and company goals, with many of those projects supporting Dropbox Dash, the company's AI-powered universal search product. Programming such as Virtual First events, ERG activities and an in-person Emerging Talent Summit rounded out the summer.

Storage, search and the path to Dash

Several projects touched the systems underneath Dropbox's products. Rhea Rai (Filesystem Data) worked on the Dropbox file history tracking system inside a large production database, citing strongly tested, verifiable code and thoughtful system design as the main lessons. The solution simplifies metadata infrastructure and cuts operational costs.

laptop computer

In Storage Core, Albert Joon Sung targeted front end latency in Magic Pocket. Elevated PUT latencies during scheduled disk restarts can delay updates in workflows such as Dash connectors, leaving users with stale or missing content. He built a cache to track storage health and added a filtering option that skips degraded volumes, reducing slow writes and giving operators more control.

sparkles

On the search side, Rishi Peddakama (Retrieval Platform) extended the unified search platform (USP), which powers search across products including Replay, to more than 20 languages. The project integrated a language detection pipeline into both indexing and retrieval, delivering multilingual search without the overhead of traditional approaches ahead of this year's Dash launch.

man technologist

Eddie Ormseth (Connector Platform) built tools giving machine learning engineers access to up-to-date information in the Dash persistence store, so downstream teams can train on fresher data and pull additional metadata fields from third-party systems without waiting for the Connector Platform team to redownload or repackage anything.

hammer and wrench

Francesca Venditti (Find & Discover) developed a document preview experience inside Dash that lets users view file content without leaving their search context. The work involved interactive UI components, PDF viewing integration and dynamic follow-up features linked to AI-powered chat.

Machine learning and AI tooling

Ben Juntilla (ML Platform) worked on a system that monitors the health of ML model deployments. AI Sentinel integrates with internal inference services to give machine learning engineers real-time operational visibility that was previously gathered manually.

Alan Zhu (Conversational AI) built an extensible AI web-automation agent, connecting Dropbox backend APIs through searchFile and uploadFile actions to fetch and upload files on open-source foundations. Keeping tool sets small and modular lets developers compose task-specific automations such as form filling or proofreading.

Developer experience and operational reliability

Ahmed Ibrahim (Web Developer Experience) worked on an AI-powered tool layered on the internal migration platform that automates code migrations. Developers can launch auto-migration jobs on selected folders for specific migration types; successful runs open a pull request automatically, and failed runs can be executed locally with changes submitted manually. The tool is customizable via the CLI or as part of an automated workflow, and Ibrahim completed two major migrations with it.

Yonatan Ginsburg (Metrics) explored advanced anomaly detection for the Vortex2 metrics system. Static alerting can miss sudden shifts that never cross predefined thresholds, or fire too often when change is expected, so he developed adaptive detection methods that adjust to evolving patterns. The approach accounts for seasonality and aims to reduce alert fatigue on call.

On the Analytics Platform team, Sanjith Udupa optimized large-scale Databricks queries and ETL pipelines for compute cost and latency. He built an optimization recommendation system that flags high-cost query patterns, expensive table-column filters and under-allocated compute resources with actionable sourcing information. He also prototyped and documented an Airflow pipeline to migrate a 500 TB mobile events log to liquid clustering.