Infrastructure and Platform Engineering

Pivot's primary cloud provider is AWS, with additional use of Cloudflare.

We deploy all our backend services to ECS (AWS's managed containers platform). Our ECS cluster uses Fargate compute. We use S3 for file/object storage. AWS Aurora PostgreSQL (using serverless and global database features) is used as our relational database system, organized into three distinct clusters (Facebox, Blockhead, and Messenger) to serve our backend services. For NoSQL workloads, we use Amazon Keyspaces (managed Cassandra).

Our web app (Expo web export), Marketing sites (Next.js), Doc site (Astro), and PivotAdmin (SSR Next.js) are deployed to Cloudflare Pages.

Quest search service uses Turbopuffer, a managed vector database, for search capabilities. We use Valkey (hosted on AWS ElastiCache Serverless) for caching across services and for Pilot's data storage.

A third-party product called Axiom is used for observability.