Poly is a 2-year-old, 6-person engineering team building a cloud storage platform where generative AI drives file discovery. The tech stack—Tauri, Electron, React, Rust, PyTorch, CUDA, Kubernetes, DynamoDB—reflects a product architecture spanning desktop clients (cross-platform via Tauri/Electron), GPU-accelerated ML inference (PyTorch + CUDA), and distributed cloud storage (Kubernetes, AWS, DynamoDB). The pain-point mix (scaling file storage, AI search performance, data pipeline throughput, global service reliability) maps directly to their project roadmap, indicating an early-stage company wrestling with the infrastructure costs and latency tradeoffs of embedding generative search at scale.
Poly builds a cloud storage platform designed for the generative-AI era, positioning AI-powered file search as the primary interaction model rather than folder hierarchies. The product ships as a desktop application (via Tauri and Electron for cross-platform reach) that connects to cloud backend services. The company is entirely engineering-focused, with mid- and senior-level IC roles, and operates from San Francisco. Active work spans AI architecture refinement, desktop OS integration, and cloud scaling—suggesting the platform is feature-complete enough to require infrastructure and performance hardening rather than foundational product work.
Poly's stack includes Tauri and Electron for the desktop client, Rust for backend systems, React and Vue for UI, PyTorch and CUDA for ML inference, and AWS (DynamoDB, CloudFront, Kubernetes) for cloud infrastructure.
Current projects include AI architecture for file search, cloud storage scaling, AI search engine development, desktop OS integration, and global service reliability—reflecting focus on performance and cross-platform consistency.
Other companies in the same industry, closest in size