Assignment$-1[$Total Marks $-25]$

M$1$: MLOps Foundations

Objective: Understand the basics of MLOps and implement a simple CI$/$CD

pipeline.

Tasks:

$1.$ Set Up a CI$/$CD Pipeline:

$$ Use a CI$/$CD tool like GitHub Actions or GitLab CI to set up a pipeline for a

sample machine learning project.

$$ Include stages for linting, testing, and deploying a simple machine

learning model.

$2.$ Version Control:

$$ Implement version control for your project using Git.

$$ Demonstrate branching, merging, and pull requests.

Deliverables:

$$ A report detailing the CI$/$CD pipeline stages.

$$ Screenshots or logs showing successful runs of the pipeline.

$$ A Git repository link with branches and merge history.

M$2$: Process and Tooling

Objective: Gain hands$-$on experience with popular MLOps tools and

understand the processes they support.

Tasks:

$1.$ Experiment Tracking:

$$ Use MLflow to track experiments for a machine learning project.

$$ Record metrics, parameters, and results of at least three different model

training runs.

$2.$ Data Versioning:

$$ Use DVC $($Data Version Control$)$ to version control a dataset used in your

project.

$$ Show how to revert to a previous version of the dataset.

Deliverables:

$$ MLflow experiment logs with different runs and their results.

$$ A DVC repository showing different versions of the dataset.

M$3$: Model Experimentation and Packaging

Objective: Train a machine learning model, perform hyperparameter tuning,

and package the model for deployment.

Tasks:

$1.$ Hyperparameter Tuning:

$$ Use a library like Optuna or Scikit$-$learn$$s GridSearchCV to perform

hyperparameter tuning on a chosen model.

$$ Document the tuning process and the best parameters found.

$2.$ Model Packaging:

$$ Package the best$-$performing model using tools like Docker and Flask.

$$ Create a Dockerfile and a simple Flask application to serve the model.

Deliverables:

$$ A report on hyperparameter tuning results.

$$ A Dockerfile and Flask application code.

$$ Screenshots of the model running in a Docker container.

M$4$: Model Deployment & Orchestration $($Optional$)$

Objective: Deploy a machine learning model and orchestrate its operations

using Kubernetes.

Tasks:

$1.$ Model Deployment:

$$ Deploy the Dockerized model from M$3$ to a cloud platform like AWS,

Azure, or GCP$.$

$$ Use a platform service like AWS ECS, Azure AKS, or Google Kubernetes

Engine $($GKE$).$

$2.$ Orchestration:

$$ Set up a Kubernetes cluster.

$$ Deploy the model using Kubernetes and create a Helm chart for

managing deployments.

Deliverables:

$$ A link to the deployed model endpoint.

$$ Kubernetes configuration files and Helm chart.

$$ A report detailing the deployment and orchestration process.

M$5$: Final Deliverables

$$ A zip file containing:

$$ Code

$$ Data

$$ Model

$$ A one$-$page summary that includes:

$$ Description of the work completed

$$ Justification for the choices made

$$ A screen recording $($maximum $5$ minutes$)$ that:

$$Explains the work done

$$Shows the results