Skip to content
MLOps Roadmap
0 / 18
Roadmap · 2026 Edition

ML
Ops.

18 stations. 3 tracks. From a notebook to reliable, monitored ML in production — at your own pace.

Foundations
~5h 0/6
Pipelines
~6h 0/6
Operations
~6h 0/6
0 of 18 stations · ~0h of ~17h
Lines —
Foundations
Pipelines
Operations
Stations —
Not started
Completed

The roadmap.

Three tracks. 18 stations. Click any node to open its detail. Mark complete as you go — your progress is saved locally.

Practice tools

Go deeper.

Interactive tools to practice what you've learned from the roadmap above.

Interactive tool

Build a CRISP prompt.

Fill in each component. Your prompt assembles live on the right.

Your prompt
Start typing in any field to see your prompt take shape…
Interactive lab

The Technique Lab.

Pick an experiment. See how each technique transforms a prompt — before and after.

Adding CRISP components to a vague prompt produces dramatically better output.

Before Vague prompt
Write a blog post about AI.
After CRISP prompt
Context: Our audience is non-technical founders who are curious about AI but intimidated by jargon.

Role: Act as a tech journalist who explains complex topics in plain English.

Instruction: Write a 600-word blog post introduction about how AI is changing content creation.

Specifics: Start with a hook anecdote. Use short paragraphs. Avoid technical terms — if you must use one, define it.

Proof: Tone example: "AI doesn't replace writers. It gives them a co-pilot that never sleeps."
AI response
Click "Simulate response" to see the difference.

    Keep reading.

    The Prompting Handbook covers the Foundation track in depth — interactive, no code required.

    Read the handbook →

    MLOps Roadmap 2026 — the full roadmap in text

    A written version of the interactive roadmap above — every station, what you'll learn, and a small thing to build — laid out for reading, reference and search.

    Foundations Start here

    F1. The ML Lifecycle

    Beginner · 45 min

    A model is never "done" — it lives in a loop: collect data, train, evaluate, deploy, monitor, and retrain. MLOps is the discipline of making that loop reliable and repeatable. Everything downstream rests on one habit: reproducibility, so any result can be recreated exactly.

    Skills: The train→deploy→monitor loop · Reproducibility · Environments vs artifacts · Why notebooks don't ship

    Build it: Take a notebook model and list every hidden dependency that would stop a teammate reproducing your result — data snapshot, random seed, library versions, hardware.

    F2. Data & Feature Versioning

    Beginner · 45 min

    Code is versioned; data usually is not — and that is where reproducibility dies. Versioning datasets and feature definitions means every model can be traced back to the exact inputs it saw, so "it worked last week" becomes a fact you can check, not a mystery.

    Skills: Dataset versioning (DVC-style) · Feature definitions · Immutable snapshots · Data lineage basics

    Build it: Snapshot a dataset, train on it, then change a few rows and re-snapshot. Prove you can reproduce the original model bit-for-bit from the first snapshot.

    F3. Experiment Tracking

    Beginner · 45 min

    Which hyperparameters produced your best model? Without tracking, that answer is lost. Experiment tracking logs params, metrics, and artifacts for every run, so you can compare hundreds of experiments and reproduce any one of them on demand.

    Skills: Logging params & metrics · Run comparison · Artifact storage · Reproducible runs

    Build it: Instrument a training script to log its hyperparameters and final metrics. Run it 5 times with different settings and rank the runs by validation score.

    F4. Model Registry & Versioning

    Intermediate · 45 min

    Once you have good models, you need a source of truth for them: which version is in staging, which is in production, what data trained it, and how to roll back. A model registry gives every model a version, a stage, and a lineage you can audit.

    Skills: Model versioning · Staging → production · Lineage & metadata · Rollback

    Build it: Design the metadata you would store for each registered model version — enough to answer "what is in production and how do I roll it back?" in one query.

    F5. Containers & Environments

    Intermediate · 45 min

    "Works on my machine" is the enemy of production ML. Containers (Docker) and pinned dependencies package your code, libraries, and runtime so training and serving behave identically on a laptop, a CI runner, and a GPU cluster.

    Skills: Docker for ML · Dependency pinning · Reproducible builds · GPU base images

    Build it: Containerize a training script with pinned dependencies. Run it on two machines and confirm the output is identical.

    F6. CI/CD for ML

    Intermediate · 60 min

    ML deserves the same automation as software — plus a few extras. A CI/CD pipeline runs tests, validates data, trains or promotes a model, and deploys it, so a change goes from commit to production without manual, error-prone steps.

    Skills: Automated tests · Data validation gates · Model promotion · Automated deploys

    Build it: Sketch a CI/CD pipeline for a model: what checks run on every commit, what gates a deploy, and what happens if data validation fails?

    Pipelines Level up

    T1. Training Pipelines & Orchestration

    Intermediate · 60 min

    A real training job is a graph of steps — ingest, validate, featurize, train, evaluate, register — that must run on a schedule, retry on failure, and be observable. Orchestrators (Airflow, Kubeflow, Prefect) turn a fragile notebook into a dependable DAG.

    Skills: DAGs & orchestration · Scheduling & retries · Step isolation · Pipeline observability

    Build it: Decompose one training script into a 5-step DAG. For each step, name its inputs, outputs, and what should happen when it fails.

    T2. Feature Stores

    Advanced · 60 min

    The classic MLOps bug: features computed one way in training and another way in serving, so the model sees different inputs in production. A feature store computes features once and serves the same values to both — killing training/serving skew.

    Skills: Training/serving skew · Online vs offline features · Feature reuse · Point-in-time correctness

    Build it: Pick a feature (e.g. "orders in last 7 days"). Explain how it could be computed differently in training vs serving, and how a feature store prevents the mismatch.

    T3. Model Packaging & Serving

    Intermediate · 60 min

    A trained model is useless until something can call it. Packaging wraps it behind a versioned API with a defined input/output contract; serving runs that API reliably, with health checks, autoscaling, and safe rollouts.

    Skills: Model as an API · I/O contracts · Health checks · Safe rollout

    Build it: Define the request/response contract for a model API. What do you return on a bad input, a timeout, or a model-not-loaded error?

    T4. Batch vs Online Inference

    Intermediate · 45 min

    Not every prediction needs to be real-time. Batch inference precomputes predictions in bulk on a schedule (cheap, high-throughput); online inference answers per request in milliseconds (costly, low-latency). Choosing right shapes your whole architecture.

    Skills: Batch scoring · Online / real-time serving · Latency vs throughput · Caching predictions

    Build it: For three use cases (recommendations email, fraud check, search ranking) decide batch or online, and justify the latency and cost tradeoff.

    T5. Scaling & GPUs

    Advanced · 60 min

    Under load, serving becomes a systems problem: autoscaling replicas, batching requests, and squeezing GPU utilization so you are not paying for idle silicon. The goal is meeting latency targets at the lowest cost per prediction.

    Skills: Autoscaling · Request batching · GPU utilization · Cost per prediction

    Build it: For a service at 1,000 requests/second with a 100ms budget, reason through replicas, batch size, and GPU count. Where is the bottleneck?

    T6. LLMOps

    Advanced · 60 min

    LLMs add a loop on top of classic MLOps: prompts are versioned artifacts, evals gate releases, and fine-tuning is its own pipeline. LLMOps is MLOps plus prompt management, eval-driven deployment, and the cost/latency realities of giant models.

    Skills: Prompt versioning · Eval-gated releases · Fine-tuning pipelines · Cost & latency control

    Build it: Sketch the extra pipeline stages an LLM feature needs that a classic model does not — prompt versioning, eval gates, guardrails — and where each sits.

    Operations Run it

    P1. Monitoring & Drift

    Advanced · 60 min

    Models rot. The world shifts under them — user behaviour changes, upstream data changes — and accuracy quietly decays with no error thrown. Monitoring watches for data drift and prediction drift so you catch decay before your users do.

    Skills: Data drift · Prediction / concept drift · Baseline distributions · Alerting on decay

    Build it: Pick a feature and define a drift check: what distribution do you compare against, and what change would trigger an alert?

    P2. A/B Testing & Rollouts

    Advanced · 60 min

    Never trust a new model blindly. Shadow it against the old one, canary it to a slice of traffic, and A/B test to prove it actually improves the metric that matters. Rollout strategy is how you ship models without betting the whole system.

    Skills: Shadow deployment · Canary releases · A/B testing · Guardrail metrics

    Build it: Design the rollout for a new ranking model: shadow → canary → full. What metric decides promotion, and what triggers an automatic rollback?

    P3. Observability & Logging

    Intermediate · 45 min

    When a prediction goes wrong at 2am, logs are all you have. Trace each request, log inputs, outputs, model version, and latency, and alert on quality and error rates. Observability turns "the model is acting weird" into a debuggable timeline.

    Skills: Request tracing · Prediction logging · Latency & error alerts · Debuggability

    Build it: Design a logging schema for a prediction service: which 8 fields do you capture, and what condition fires an alert? Write it as JSON.

    P4. Cost & Resource Optimization

    Intermediate · 45 min

    ML infrastructure bills add up fast — idle GPUs, oversized instances, redundant recomputation. Right-sizing compute, caching predictions, spot instances, and turning off what you are not using keeps the system affordable at scale.

    Skills: Right-sizing compute · Caching & spot instances · Idle-resource cleanup · Cost attribution

    Build it: Audit a serving setup for waste: where would you cache, where would you downsize, and where is a GPU sitting idle you could reclaim?

    P5. Governance & Lineage

    Intermediate · 45 min

    In regulated or high-stakes settings you must answer: what data trained this model, who approved it, and why did it make that decision? Lineage and governance track the full provenance of every model and prediction for audit and compliance.

    Skills: Model & data lineage · Audit trails · Access control · Compliance & documentation

    Build it: For a model in a regulated domain, list what you would need to store to answer an auditor: training data, approvals, and decision traceability.

    P6. Incident Response & Retraining

    Advanced · 60 min

    The loop closes here. When monitoring flags a regression, you need a runbook: roll back to a known-good version, diagnose the cause, and trigger retraining on fresh data. Automated retraining pipelines turn decay from a crisis into a routine.

    Skills: Rollback runbooks · Root-cause diagnosis · Automated retraining · Closing the loop

    Build it: Write an incident runbook: a model's accuracy drops 10% overnight — what are your first five steps, in order?

    Finished this one? 0 / 13 Roadmaps done

    Explore the topic

    See this alongside everything else on the same subject — handbooks, system designs, challenges and tools, in one place.

    More Roadmaps