series

100 Days of MLOps

A beginner-first, project-based path that takes you from "what is MLOps?" to running an end-to-end machine-learning platform on your own laptop. No computer-science background required and no cloud bills ever: every day explains the idea in plain English, builds one small working project, and runs 100% locally on macOS, Linux or Windows. You start with a clean environment and real ML, then layer on reproducibility and data/model versioning (Git + DVC), experiment tracking (MLflow), data validation and feature stores, model packaging and serving (FastAPI, Docker, BentoML), pipeline orchestration (Prefect), CI/CD for ML, Kubernetes deployment on a local cluster (kind), and finally monitoring, drift detection and automated retraining — capped by a capstone that ties all ten modules into one runnable platform.

90/100 days published ~14h total Absolute Beginner → MLOps Engineer
–

Your progress starts here

90 published · ~14h of hands-on builds · sign in to sync progress across devices

01Foundations & Your Local MLOps Lab

What MLOps really is, a bulletproof cross-OS setup, Python environments and Git for ML, your first end-to-end model, clean project structure, and reproducibility from day one.

0/10
  1. D01What is MLOps, and Setting Up Your Machine on Any OS19 min read
  2. D02The MLOps Lifecycle & Mental Model17 min read
  3. D03Python Environments for ML: venv, pip & requirements.txt13 min read
  4. D04Git Basics for ML Projects (and a Proper .gitignore)12 min read
  5. D05Notebooks vs Scripts: Explore Fast, Ship Reliably13 min read
  6. D06Your First ML Model, End to End13 min read
  7. D07Saving & Loading Models with joblib10 min read
  8. D08Project Structure That Scales11 min read
  9. D09Task Automation, Cross-OS: Makefile & invoke10 min read
  10. D10Reproducibility 101: Same Result, Every Time10 min read

02Machine Learning You Can Operationalize

Just enough honest, practical ML to have real models worth shipping — data handling, leak-free splits, metrics, sklearn pipelines, cross-validation, tuning and packaging a training run.

0/10
  1. D11Working with Data: pandas for ML12 min read
  2. D12Train, Validation & Test Splits (and Data Leakage)11 min read
  3. D13Classification Models & Metrics10 min read
  4. D14Regression Models & Metrics: MAE, RMSE & R²10 min read
  5. D15Feature Engineering with scikit-learn Pipelines10 min read
  6. D16Cross-Validation & Honest Evaluation10 min read
  7. D17Hyperparameters & Tuning with GridSearchCV9 min read
  8. D18Handling Imbalanced & Messy Data9 min read
  9. D19Model Interpretability Basics9 min read
  10. D20Packaging a Training Run9 min read

03Reproducibility & Versioning: Data + Code

Version everything with Git and DVC — datasets, pipelines, params and models — so any result can be reproduced from a clean clone.

0/10
  1. D21Why Data Versioning? The Problem10 min read
  2. D22Intro to DVC: Version Your Data Like Code8 min read
  3. D23DVC Remotes: Sharing Versioned Data8 min read
  4. D24DVC Pipelines: Reproducible ML Workflows8 min read
  5. D25Reproducing & Comparing Runs with DVC8 min read
  6. D26Config Management with YAML & Hydra8 min read
  7. D27Environment Reproducibility: Lockfiles & Docker9 min read
  8. D28Versioned Data & Model Artifacts8 min read
  9. D29Project Templates: Scaffold Every Project the Right Way7 min read
  10. D30Capstone: A Fully Reproducible Project8 min read

04Experiment Tracking with MLflow

Never lose a result again: track params, metrics and artifacts, compare runs, autolog, tune with Optuna, and run a local tracking server.

0/10
  1. D31Why Experiment Tracking? The Spreadsheet of Doom9 min read
  2. D32MLflow Tracking Basics7 min read
  3. D33Comparing Runs in the MLflow UI8 min read
  4. D34MLflow Autologging7 min read
  5. D35Logging Models & Artifacts with Signatures8 min read
  6. D36Hyperparameter Tuning, Tracked: Optuna + MLflow8 min read
  7. D37MLflow Projects: Make Training Runnable Anywhere8 min read
  8. D38Organizing Experiments: Names, Tags & Nested Runs7 min read
  9. D39A Local MLflow Tracking Server8 min read
  10. D40Capstone: An Experimentation Workflow8 min read

05Data Quality, Validation & Feature Stores

Catch bad data before it breaks models with Pandera and Great Expectations, profile and document datasets, and serve consistent features with Feast.

0/10
  1. D41Why Data Validation Matters8 min read
  2. D42Schema Validation with Pandera8 min read
  3. D43Data Validation with Great Expectations8 min read
  4. D44Validation as a Pipeline Gate8 min read
  5. D45Data Profiling & Documentation8 min read
  6. D46Reusable Feature Pipelines9 min read
  7. D47Intro to Feature Stores with Feast9 min read
  8. D48Online vs Offline Features7 min read
  9. D49Preventing Training/Serving Skew8 min read
  10. D50Capstone: A Validated Feature Pipeline8 min read

06Packaging & Serving Models

Turn a model into a real service — FastAPI, Pydantic validation, Docker, tests, batch vs online, BentoML, load testing and ONNX.

0/10
  1. D51From Model to Inference API8 min read
  2. D52Serving a Model with FastAPI7 min read
  3. D53Request/Response Validation with Pydantic7 min read
  4. D54Dockerizing Your Model Service7 min read
  5. D55Testing Your Model API8 min read
  6. D56Batch vs Online Inference8 min read
  7. D57Model Serving with BentoML8 min read
  8. D58Latency & Load Testing with Locust7 min read
  9. D59Optimizing Models with ONNX8 min read
  10. D60Capstone: A Production-Style Model Service7 min read

07Orchestration & Automated Pipelines

Pipelines that run themselves with Prefect — retries, caching, scheduling, parameterization and an end-to-end training pipeline (plus Airflow awareness).

0/10
  1. D61Why Orchestration? Cron Isn't Enough8 min read
  2. D62Intro to Prefect10 min read
  3. D63Retries, Caching & Logging10 min read
  4. D64Scheduling Pipelines9 min read
  5. D65Parameterized Pipelines9 min read
  6. D66Complex Flows: Subflows & Mapping10 min read
  7. D67Observability: The Prefect UI & Run States9 min read
  8. D68A Peek at Airflow10 min read
  9. D69Notifications & Failure Handling9 min read
  10. D70Capstone: An Automated Retraining Pipeline11 min read

08CI/CD for Machine Learning

Ship model changes safely: test ML code and models, GitHub Actions, continuous training, CML reports, validation gates and registry promotion.

0/10
  1. D71Why Monitor? Models Decay10 min read
  2. D72Logging Predictions & Ground Truth10 min read
  3. D73Detecting Data Drift (PSI & KS)9 min read
  4. D74Concept Drift & Performance Decay10 min read
  5. D75Intro to Evidently8 min read
  6. D76Drift Dashboards & Reports9 min read
  7. D77Monitoring Metrics (Prometheus-style)9 min read
  8. D78Alerting on Drift10 min read
  9. D79Closing the Loop: Monitoring-Triggered Retraining10 min read
  10. D80Capstone: A Fully Monitored ML System10 min read

09Deploying Models on Kubernetes, Locally

Real serving infra on a local kind cluster — deploy, configure, scale, KServe/Seldon, canary and A/B, rollbacks and Argo CD GitOps.

0/10
  1. D81Why Kubernetes? Deploying at Scale9 min read
  2. D82Kubernetes Core Concepts9 min read
  3. D83Deploying Your Model Service8 min read
  4. D84Services & Networking8 min read
  5. D85Scaling & Autoscaling8 min read
  6. D86Rolling Updates & Rollbacks9 min read
  7. D87Config & Secrets8 min read
  8. D88Health Checks & Resource Management9 min read
  9. D89Packaging with Helm8 min read
  10. D90Capstone: Deploying the Model Service at Scale8 min read

10Monitoring, Drift & the Full MLOps Loop

Models rot — detect it and respond: Prometheus/Grafana, prediction logging, data and concept drift with Evidently, alerting, automated retraining, governance, and the final capstone.

0/0

    11Coming up

    One new day at a time — follow @syssignals to catch each release.

    1. D91Why Monitoring ML Is Differentsoon
    2. D92Service Monitoring with Prometheus & Grafanasoon
    3. D93Logging Predictions & Ground Truthsoon
    4. D94Data Drift Detection with Evidentlysoon
    5. D95Concept Drift & Performance Decaysoon
    6. D96Alerting on Drift & Degradationsoon
    7. D97Automated Retraining Triggerssoon
    8. D98Governance, Lineage & Model Cardssoon
    9. D99The Full MLOps Architecturesoon
    10. D100Capstone: Your End-to-End Local MLOps Platformsoon