Skip to content
Datamewby Intura
Start free
MLOps: Deploying Machine Learning & Deep Learning Models

Serving Deep Learning Models and LLMs

GPUs, batching, quantisation and formats like ONNX — making large models fast enough and cheap enough.

Level
Intermediate
Lessons
4
Total time
1h 6m
Simulations
—
Included in MLOps: Deploying Machine Learning & Deep Learning Models · $9
Tools you'll use
  • Python
  • Docker
  • FastAPI
  • Kubernetes
  • GitHub
Skills
  • ONNX
  • Quantisation
  • Dynamic batching

What you will be able to do

  • Trade off latency, throughput and GPU cost
  • Shrink a model without meaningful quality loss

Before you start

A suggestion, never a lock — you can start here.

  • packaging-and-serving-models
Modules that help first

Syllabus

4 lessons · 1h 6m

Every lesson's title and summary is public. The lessons themselves open once you own the track.

Every lesson in MLOps: Deploying Machine Learning & Deep Learning Models

  1. Latency, Throughput and Batching

    16 min

    Why bigger batches raise throughput and latency at the same time, where the saturation point is, and the two knobs that control it in an inference server.

    Included with the track

  2. ONNX and Model Optimisation

    16 min

    Exporting a model to one graph format that a C++ runtime executes on CPU, GPU and phones — and verifying that the result is still the same model.

    Included with the track

  3. Quantisation

    16 min

    int8 and 4-bit: how float weights are mapped to small integers, what you lose, what you gain, and how to measure it on your own model.

    Included with the track

  4. Serving LLMs

    18 min

    The KV cache, continuous batching and inference servers: why serving a language model is a GPU-memory problem, and which knobs decide how many users one card can serve.

    Included with the track
Lifetime access
$9
Buy the track
More in MLOps: Deploying Machine Learning & Deep Learning Models
Module · Intermediate

ML Systems in Production

Why models degrade after launch, how batch and online inference differ, and what to monitor.

  • 4 lessons
  • 1h 4m
  • 4 simulations
Included in MLOps: Deploying Machine Learning & Deep Learning Models· $9
Module · Intermediate

Packaging and Serving Models

From notebook to service: saving models, an API with FastAPI, and reproducible Docker containers.

  • 3 lessons
  • 48 min
Included in MLOps: Deploying Machine Learning & Deep Learning Models· $9
Module · Intermediate

ML Pipelines and CI/CD

Experiment tracking, model registries, and automated pipelines from raw data to a released model.

  • 3 lessons
  • 47 min
  • 2 simulations
Included in MLOps: Deploying Machine Learning & Deep Learning Models· $9
Module · Intermediate

Monitoring and Retraining

Drift detection, performance with delayed labels, retraining strategies, and safe rollouts.

  • 3 lessons
  • 48 min
  • 4 simulations
Included in MLOps: Deploying Machine Learning & Deep Learning Models· $9