Loading

🎥Video 25 min1 / 5

What is MLOps?

What is MLOps?

MLOps (Machine Learning Operations) is the set of practices that bridges the gap between developing ML models and running them in production.

The Problem MLOps Solves:

Without MLOps, this happens frequently:

Data Scientist builds a model with 94% accuracy on their machine
Tries to move it to production → takes weeks or fails
After deployment nobody monitors it → performance degrades without anyone knowing

Core Components:

Data Management — data versioning and management (DVC)
Experiment Tracking — tracking experiments and parameters (MLflow)
Model Registry — storing and versioning models
CI/CD for ML — automating testing and deployment
Monitoring — monitoring performance and drift in production

Maturity Levels:

Level 0 — everything manual, no automation
Level 1 — experiment and model tracking
Level 2 — automated CI/CD
Level 3 — full monitoring and automatic retraining

python

from dataclasses import dataclass, field
from typing import List, Tuple

# ─── MLOps Maturity Model ──────────────────────────────────
@dataclass
class MLProject:
    name:               str
    model_type:         str
    version:            str   = "1.0.0"
    accuracy:           float = 0.0
    is_deployed:        bool  = False
    tracking_enabled:   bool  = False
    ci_cd_enabled:      bool  = False
    monitoring_enabled: bool  = False

    def mlops_score(self) -> int:
        return sum([
            25 if self.tracking_enabled   else 0,
            25 if self.is_deployed        else 0,
            25 if self.ci_cd_enabled      else 0,
            25 if self.monitoring_enabled else 0,
        ])

    def maturity_level(self) -> str:
        s = self.mlops_score()
        if s == 0:  return "Level 0 — Manual (لا أتمتة)"
        if s <= 25: return "Level 1 — Tracking"
        if s <= 50: return "Level 2 — Deployed"
        if s <= 75: return "Level 3 — CI/CD"
        return       "Level 4 — Full MLOps ✨"

    def report(self):
        score = self.mlops_score()
        level = self.maturity_level()
        print(f"\n📊 {self.name} (v{self.version})")
        print(f"   النوع    : {self.model_type}")
        print(f"   Accuracy : {self.accuracy:.1%}")
        print(f"   Score    : {score}/100 — {level}")
        checks: List[Tuple[str, bool]] = [
            ("Experiment Tracking", self.tracking_enabled),
            ("Deployed to Prod",    self.is_deployed),
            ("CI/CD Automated",     self.ci_cd_enabled),
            ("Monitoring Active",   self.monitoring_enabled),
        ]
        for label, ok in checks:
            icon = "✅" if ok else "❌"
            print(f"   {icon} {label}")

# ─── قبل MLOps ─────────────────────────────────────────────
print("🔬 الوضع بدون MLOps:")
print("-" * 50)
proto = MLProject(
    name="sentiment-model", model_type="Text Classifier",
    version="1.0.0", accuracy=0.89,
)
proto.report()

# ─── بعد MLOps ─────────────────────────────────────────────
print(f"\n🚀 الوضع بعد تطبيق MLOps:")
print("-" * 50)
prod = MLProject(
    name="sentiment-model", model_type="Text Classifier",
    version="2.3.1", accuracy=0.94,
    is_deployed=True, tracking_enabled=True,
    ci_cd_enabled=True, monitoring_enabled=True,
)
prod.report()

# ─── مكونات MLOps ──────────────────────────────────────────
print(f"\n\n🧱 مكونات MLOps الرئيسية:")
components = [
    ("Data Management",  "DVC, Delta Lake",        "إدارة وإصدار البيانات"),
    ("Experiment Track", "MLflow, W&B",             "تتبع التجارب والمعاملات"),
    ("Model Registry",   "MLflow, HuggingFace Hub", "تخزين وإصدار النماذج"),
    ("CI/CD",            "GitHub Actions, Jenkins", "أتمتة الاختبار والنشر"),
    ("Serving",          "FastAPI, TorchServe",     "خدمة النموذج للإنتاج"),
    ("Monitoring",       "Evidently, Grafana",      "مراقبة الجودة والانجراف"),
]
for comp, tools, desc in components:
    print(f"  📦 {comp:<20} — {desc}")
    print(f"     أدوات: {tools}")
    print()