Pometry offers an expansive library of 50+ algorithms, including centrality, community detection (static and temporal), pathfinding, graph embeddings (e.g., FastRP), link prediction, and specialised temporal algorithms like temporal SIR (for spread) and Motif (for behaviours).

You can define an algorithm once and then run it over different time windows (rolling, specific snapshots), or against data using discrete event times or continuous time models. This allows you to analyse how graph structures and algorithm results change and evolve.

Incredibly easy. You can write custom graph algorithms in Python or Rust using a simple, consistent API, often in just seconds. These custom kernels introduce no performance overhead and can be directly exposed via GraphQL for application use.

Our platform is data source agnostic. We can run analytics directly on data in various formats like Parquet, CSV, our native store, and connect to data lakes, warehouses like Snowflake, or federated AWS data sources.

Our platform is engineered at the forefront of innovation. Techniques like vectorised execution, SIMD operations, leveraging modern CPU instruction sets like AVX512, combined with efficient memory management and parallelism allow us to deliver our benchmarked speed and scale.

Absolutely. Pometry is designed for massive scale. Our architecture handles parallelism efficiently and can scale to terabytes of data on a single instance by decoupling compute from storage, allowing you to focus on your analytical questions, not on complex infrastructure.

Yes! Our graph system is a tiny 10MB binary. This means you can run it embedded (in-process) directly within your Python applications, on-prem, cloud, air-gapped, or at the edge. This is ideal for scenarios requiring ultra-low latency and tight integration with ML pipelines.