From Small Bets to Scaled Systems: Operationalizing Contextual, Multimodal AI for Productivity

Ali will share how Dropbox tackled multimodal and contextual AI enterprise challenges. Using Dropbox and Dropbox Dash as an example, Ali will dive into how Dropbox built the large-scale AI infrastructure required to index, retrieve, and reason over massive volumes of multimodal content. He’ll explore how media indexing and retrieval challenges were addressed at the infrastructure layer to enable reliable, low-latency search and chat experiences at enterprise scale, and what it takes to make multimodal intelligence practical in a real-world product.He'll also cover key engineering trade-offs across data ingestion, indexing, and serving, as well as the data center and compute decisions required to make inference fast, predictable, and cost-efficient. Highlights will also include how small efficiency choices in batching, caching, scheduling, and data movement compound into significant gains in throughput, tail latency, and cost—making AI viable in production, not just in demos.