DDrona4U
Sign inCreate account
My learningMLOPS TutorialLesson 19
Prompts as Code: Versioning, Evals & A/B Testing the New Model Artifact

Lesson 19

Prompts as Code: Versioning, Evals & A/B Testing the New Model Artifact

In the era of generative AI, prompts have become as important as models themselves. In this video, explore how treating prompts as code enables version control, automated evaluations, experimentation, and reproducible deployments. Learn how leading AI teams manage prompt changes, run A/B tests, track performance, and turn prompts into first-class artifacts within modern MLOps and LLMOps workflows.

Get the full lesson

Sign in to unlock everything beyond the preview — it's free.

  • Take timestamped notes as you watch
  • Read the full transcript and download resources
  • Join the discussion and track your progress
Sign inCreate free account

Curriculum

22 lessons · 1h 36m

0/22 lessons done1h 36m left
  1. 0103:50

    From Notebook Chaos to Autonomous Retraining: The 4 Levels of MLOps Maturity

    03:50

  2. 0204:26

    The Reproducibility Receipt: Proof Your ML Results Can Be Trusted

    04:26

  3. 0304:07

    MLflow 3.x: One Registry for XGBoost, GPTs, and Everything Between

    04:07

  4. 0404:24

    Git for Data: DVC, lakeFS, and the End of “Works on My Laptop

    04:24

  5. 0503:58

    When MLflow Stops Scaling: Choosing Between W&B, Comet, and Neptune

    03:58

  6. 0603:49

    Training-Serving Skew: The Silent Model Killer at 3 A.M

    03:49

  7. 0703:16

    feast vs tecton vs hopsworks — the feature store choice you'll live with for 5 years

    03:16

  8. 0805:18

    The Sub-10ms Inference Stack: How Redis & Kafka Deliver Features at Lightning Speed

    05:18

  9. 0903:59

    The Orchestrator Wars: Kubeflow vs Airflow vs Prefect vs Flyte

    03:59

  10. 1004:13

    CI/CD for Models: GitHub Actions That Train, Test, and Block Bad Deploys

    04:13

  11. 1105:35

    The Death of Nightly Retraining: Drift-Aware, Event-Driven ML Pipelines

    05:35

  12. 1203:30

    Batch vs Online vs Streaming Inference: The Choice That Shapes Your Cloud Bill

    03:30

  13. 1304:14

    The Serving Stack Showdown: BentoML vs KServe vs Seldon vs Triton

    04:14

  14. 1404:36

    Shadow Traffic, Canaries & Ramp-Ups: Deploy Models Without Breaking Production

    04:36

  15. 1504:28

    100 Models, One GPU, Zero Meltdowns: The Multi-Model Endpoint Pattern

    04:28

  16. 1603:25

    Three Flavors of Drift: Data, Concept & Prediction—and How to Fix Each One

    03:25

  17. 1705:31

    The Silent Failure Stack: Evidently vs Arize vs Fiddler vs WhyLabs

    05:31

  18. 1804:27

    Model Cards, Lineage Logs & the 2 A.M. Audit Call You Didn’t See Coming

    04:27

  19. 04:27

    Prompts as Code: Versioning, Evals & A/B Testing the New Model Artifact

    04:27

  20. 2004:48

    RAG in Production: The Chunking, Embedding & Vector DB Decisions That Make or Break Recall

    04:48

  21. 2106:13

    Catching Hallucinations Live: The Guardrail Stack That Keeps LLMs Honest

    06:13

  22. 2203:08

    AgentOps & Trace Stores: Debugging the Agent That Fails on Step Seven

    03:08