RedisAI is a module bringing tensors to the Redis database and the ability to run computations on them. This brings the operational flexibility and the performance of Redis into the landscape of AI serving. By keeping the data local to the computation, roundtrip is zero, inference happens right where the data resides. RedisAI supports HA, Cluster and Redis Enterprise.
RedisAI supports the most popular inference engines: PyTorch, TensorFlow, ONNX and ONNX-ML. Whatever the framework is, RedisAI exposes a uniform API to developers. Changing the backend has an extremely limited impact on a codebase. RedisAI fits all scales, from edge to web-scale, from small ARM devices to multi-node clusters on the cloud.
RedisAI is not your usual serving solution. Build DAGs of RedisAI commands for maximum performance. Or programmatically compose computations using TorchScript, managing ensambles, model aliases, A/B testing, monitoring metrics, uncertainty estimation. Optimize calls with automated batching, leverage clustering by running commands on nodes or replicas. Compose away.
RedisAI is a module bringing tensors to the Redis database and the ability to run computations on them. This brings the operational flexibility and the performance of Redis into the landscape of AI serving. By keeping the data local to the computation, roundtrip is zero, inference happens right where the data resides. RedisAI supports HA, Cluster and Redis Enterprise.
RedisAI supports the most popular inference engines: PyTorch, TensorFlow, ONNX and ONNX-ML. Whatever the framework is, RedisAI exposes a uniform API to developers. Changing the backend has an extremely limited impact on a codebase. RedisAI fits all scales, from edge to web-scale, from small ARM devices to multi-node clusters on the cloud.
RedisAI is not your usual serving solution. Build DAGs of RedisAI commands for maximum performance. Or programmatically compose computations using TorchScript, managing ensambles, model aliases, A/B testing, monitoring metrics, uncertainty estimation. Optimize calls with automated batching, leverage clustering by running commands on nodes or replicas. Compose away.
Deploy models from TensorFlow, PyTorch, ONNX, ONNX-ML. Run TorchScript natively for model composition.
Dramatically accelerate & simplify AI serving by bringing model execution to where the data lives.
Leverage Redis master-replica and cluster capabilities to design your compute architecture.
Use RedisAI as a deployment target for your lifecycle management framework of choice.
