Skip to content

Why Senren?

The Problem: ML Infrastructure is Hard to Manage

Modern ML applications require complex serving infrastructure:

  • Feature stores (online + offline) for real-time inference
  • Document stores and search indexes for ranking pipelines
  • Embeddings storage and semantic search capabilities
  • Model servers for inference across multiple models
  • A/B testing infrastructure with traffic routing
  • Data pipelines connecting production → warehouses → training datasets

The Challenges

1. Multi-cloud complexity

Your application runs on AWS, but your data team uses GCP. You need infrastructure in both clouds, synchronized and consistent.

2. Multi-region deployments

Low-latency inference requires infrastructure close to users. You need the same Redis feature store in us-east-1, eu-west-1, and asia-east-1.

3. Multi-cluster coordination

Multiple Kubernetes clusters in the same region (for isolation, experimentation, or shadow traffic testing) all need coordinated infrastructure provisioning.

4. Declarative management

Managing infrastructure imperatively doesn't scale:

  • "Create Redis in us-central"
  • "Oh wait, also create it in eu-west"
  • "Actually, increase memory in us-central"
  • "Did I already do eu-west?"

You want to declare desired state and let the system reconcile.

5. Consistency across environments

Feature stores must be identical across regions (schema, configuration). Manual management leads to drift and subtle bugs in production.

What Senren Provides

Unified control plane for ML infrastructure across clouds, regions, and clusters.

Define your infrastructure once using Python:

from senren import SenrenClient, Database

client = SenrenClient(endpoint="control.senren.dev")

# Deploy feature store across AWS + GCP, multiple regions
client.apply_state(
    databases=[
        Database(
            name="user-features",
            type="redis-cluster",
            memory_mb=8192,
            regions=[
                "aws:us-east-1",
                "aws:eu-west-1",
                "gcp:us-central1",
                "gcp:asia-east1",
            ],
        )
    ]
)

Senren automatically:

  • ✅ Provisions infrastructure in each cloud/region
  • ✅ Synchronizes configuration across all locations
  • ✅ Handles failures and retries
  • ✅ Reports unified status back to control plane

Core Promises

  1. Multi-cloud native: Deploy to AWS, GCP (Azure coming) from a single API
  2. Declarative state management: Define desired state, Senren reconciles
  3. Regional autonomy: Regional planes operate independently (no central bottleneck)
  4. Built for ML workloads: Feature stores, search indexes, model serving (not just databases)
  5. Python-first API: Integrate with your ML pipelines naturally

Next Steps