AI

ESIA

School of Artificial Intelligence

Worldwide cohort
Students across time zones
HomeMaster’sTrainingsProjectsResearchBlogAboutContact

Master’s Program

MLOps & Production AI

Deploy, monitor and iterate on real AI systems

Overview

This track is about turning models into production systems. You learn containerization, CI/CD basics, monitoring signals (performance/drift), and safe iteration practices with versioned releases.

Key outcomes

  • Containerize and deploy model services
  • Track experiments and model versions
  • Define monitoring signals and retraining triggers
  • Adopt a production-ready workflow with CI/CD fundamentals

Format

  • Engineering-oriented labs
  • Remote-friendly (worldwide cohorts)

Tools

  • Docker
  • FastAPI
  • GitHub Actions basics
  • MLflow basics

Detailed program

Production-first MLOps: ship model services with versioning, CI/CD basics, monitoring signals and safe iteration practices used by real teams.

Module 1 — Production Foundations (Reproducibility & Packaging)

Turn notebooks into reliable, reproducible codebases.

1–2 weeks

What you’ll learn

  • Project structure: src/, configs, environments, reproducibility
  • Dependency management and lockfiles (why they matter)
  • Training vs inference parity (same transforms, same schemas)
  • Testing basics: unit tests for transforms and validation

Skills you’ll gain

Reproducible ML projectsPackaging disciplineTrain/inference parityTesting mindset

Module 2 — Model Serving (FastAPI patterns)

Serve models like products: schemas, validation, stability.

2–3 weeks

What you’ll learn

  • FastAPI endpoints: predict, health, version info
  • Input validation and payload schemas (prevent bad requests)
  • Performance basics: batching awareness, serialization, caching basics
  • Observability basics: logs, metrics, request tracing mindset

Skills you’ll gain

Model serving with FastAPIInput validationService stability mindsetBasic observability

Module 3 — Containerization (Docker for ML services)

Package your service to run the same everywhere.

2 weeks

What you’ll learn

  • Dockerfile best practices (small images, reproducibility)
  • Environment configuration (dev/prod parity, secrets awareness)
  • Local orchestration basics (compose mindset if needed)
  • Build/run/debug workflow

Skills you’ll gain

Dockerized ML servicesEnvironment managementDebugging containersDeployment readiness

Module 4 — Experiment Tracking & Model Versioning (MLflow basics)

Track runs, artifacts and manage model versions correctly.

2 weeks

What you’ll learn

  • Tracking runs: params, metrics, artifacts
  • Model packaging: signature, input schema, reproducibility
  • Versioning discipline: tags, stages, release notes mindset
  • Rollback mindset and reproducible retraining

Skills you’ll gain

Experiment trackingModel versioning basicsRelease disciplineRollback thinking

Module 5 — CI/CD Fundamentals for ML

Automate checks so you can ship safely.

1–2 weeks

What you’ll learn

  • Linting + tests + basic quality gates
  • Build pipeline basics (Docker build, push, versioning)
  • Deployment triggers and environment separation mindset
  • Model checks: schema checks, basic regression tests

Skills you’ll gain

CI/CD mindsetQuality gatesAutomated buildsSafe release workflow

Module 6 — Monitoring, Drift & Iteration

Know when a model degrades and what to do next.

2 weeks

What you’ll learn

  • Monitoring signals: latency, error rates, data drift, performance proxies
  • Drift basics: what it is, why it happens, what to track
  • Retraining triggers and feedback loop design
  • Incident response basics: triage, rollback, mitigation plan

Skills you’ll gain

Monitoring strategyDrift awarenessIteration & retraining planningIncident response mindset

Capstone — Production-ready ML API

Deliver a deployable ML service with versioning, tests and monitoring plan.

2–3 weeks

What you’ll learn

  • Package a trained model + API + Docker image
  • Add tests + basic CI pipeline
  • Document deployment + monitoring signals
  • Deliver a demo and technical README

Skills you’ll gain

End-to-end MLOps deliveryDeployable serviceDocumentation & demoProduction discipline