Deploy Your AI Models to Production

You will learn to:

• Work with vectors, matrices, and embeddings in an AI model context.
• Understand the role of embeddings for similarity, semantic search, and RAG.
• Use statistics to track input data quality and prediction quality.
• Lay the first monitoring building blocks: distributions, thresholds, simple anomaly detection.

You will learn to:

• Build data pipelines dedicated to Machine Learning (batch & near real‑time).
• Design and document reusable features for multiple models.
• Set up a Feature Store to centralize and share features.
• Version data, features, and pipelines with DVC / Git to ensure traceability.

You will learn to:

• Train regression and classification models with Scikit‑Learn.
• Structure your experiments (train / validation / test, cross‑validation).
• Choose and interpret the right metrics (AUC, F1, precision, recall, etc.).
• Prepare your models for production: serialization, signatures, dependencies.

You will learn to:

• Understand what Transformers bring to NLP and other tasks (vision, multimodal).
• Use the Hugging Face Hub to select, test, and version pre‑trained models.
• Apply Transfer Learning on concrete cases (classification, Q&A…).
• Integrate these models into an existing MLOps stack (tracking, serving, monitoring).

You will learn to:

• Instrument your experiments with MLflow (params, metrics, artifacts, models).
• Deploy an MLflow Tracking Server and structure the team's experiment space.
• Use the Model Registry to manage model versions (Staging, Production, Archive).
• Manage the full model lifecycle: promotion, rollback, audit, reproducibility.

You will learn to:

• Expose an AI model via a robust API with FastAPI & Pydantic.
• Design and validate ML endpoint input / output schemas.
• Dockerize a scoring API and prepare deployment-ready images.
• Structure AI micro‑services and integrate them into an existing architecture.

You will learn to:

• Design robust prompts for concrete business cases.
• Set up an end-to-end RAG pipeline: ingestion, chunking, embeddings, indexing.
• Use a vector database (Qdrant) for semantic search.
• Orchestrate your GenAI application with LangChain and explore Agents.

You will learn to:

• Set up model tracking metrics (SLI / SLO, predictive quality, latency).
• Detect and analyze Data Drift and Concept Drift.
• Use Evidently AI to monitor models over time.
• Build supervision dashboards with Grafana for the Data / Product team.

You will:

• Design a complete architecture combining a classification API & RAG Chatbot.
• Develop a support ticket classification API and put it into production.
• Build a RAG chatbot connected to your document base (FAQs, PDFs, historical tickets).
• Integrate tracking, versioning, and monitoring to deliver an AI product usable by a team.