PyStreamAI

Simple ML Deployment - Kubeflow + KServe Replacement

Zero YAML. No Kubernetes expertise required. Ship models to production in minutes.

Vision

Kubeflow solved the problem of running ML workloads on Kubernetes, but accumulated complexity. KServe requires infrastructure expertise. Modern ML engineers shouldn't need to become Kubernetes experts to deploy models.

PyStreamAI is designed to work without Kubernetes by default—it runs locally with Docker, scales to cloud (AWS/GCP/Azure), and optionally uses Kubernetes if you already have it.

PyStreamAI replaces both Kubeflow + KServe with:

• ✨ Dead-simple Python API — No YAML, no manifests, no infrastructure expertise
• 🚀 Local-to-cloud — Test locally, deploy with one command
• 💰 Cost visibility — Real-time cost attribution per request
• 🌍 Multi-cloud — AWS, GCP, Azure, on-prem, Kubernetes (your choice)
• 🎯 Modern AI stack — LLMs, RAG, agents, not just classical ML
• ⚡ Performance — Rust core for speed + Python ergonomics

Quick Start

Installation

pip install pystreamai

Simple Deployment

from pystreamai import Platform

platform = Platform()

# Train a model
@platform.train(gpu="A100", time_limit="1h")
def train_model(data):
    model = train_bert(data)
    return model

job = train_model(data="s3://bucket/data")

# Deploy it
@platform.serve(replicas=3, gpu="L4")
def predict(x):
    return model.predict(x)

endpoint = predict()

# Use it
result = endpoint.predict({"text": "Hello world"})

That's it. No YAML. No Kubernetes. No infrastructure knowledge required.

Architecture

┌──────────────────────────────────────┐
│   PyStreamAI Python API              │
├──────────────────────────────────────┤
│  @train  @serve  @pipeline           │
├──────────────────────────────────────┤
│   Rust Core Orchestrator             │
├──────────────────────────────────────┤
│ Scheduler │ Executor │ Storage       │
├──────────────────────────────────────┤
│         Backend Abstraction           │
└──────────────────────────────────────┘
     ↓ Deploy to ↓
Local/Docker → AWS → GCP → Azure → Kubernetes (optional)

Backend Options (choose what works for you):

• Local (default) — Develop locally with Docker
• Docker — Run containerized models
• AWS — Lambda + EC2 + SageMaker integration
• GCP — Cloud Run + Compute Engine
• Azure — Container Instances + AML
• Kubernetes — If you already use K8s (but not required)

Features (v0.1)

• Simple Python API
• Local deployment
• Training job submission
• Model serving
• Cost tracking
• Multi-cloud backends
• Observability (traces, metrics, logs)
• Pipelines (DAGs + dynamic)
• Model registry
• Edge deployment

Development

Setup

# Install Rust
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh

# Install Python dependencies
pip install maturin pytest

# Build
maturin develop

Running Tests

pytest tests/

Roadmap

v0.1 (current)

• Simple Python API
• Local & AWS backend
• Training + serving

v0.2

• Pipelines (workflows)
• Multi-cloud (GCP, Azure)
• Cost tracking

v0.3

• Model registry
• Observability layer
• Prompt management

v0.4+

• RAG workflows
• Agent framework
• Edge deployment

License

Proprietary. © 2026 Georgi Mammen Mullassery.

Status

🚧 Active development. Not ready for production yet.