ML lifecycle stages in MLOps - Step-by-Step Execution

Process Flow - ML lifecycle stages

Data Collection

↓

Data Preparation

↓

Model Training

↓

Model Evaluation

↓

Model Deployment

↓

Monitoring & Maintenance

↩Back to Data Collection (if needed)

The ML lifecycle moves step-by-step from gathering data to preparing it, training a model, checking its quality, deploying it, and then monitoring it for improvements.

Execution Sample

MLOps

Stage = 'Data Collection'
if Stage == 'Data Collection':
    Stage = 'Data Preparation'
# ... continues through stages

This code simulates moving through the ML lifecycle stages one by one.

Process Table

Step	Current Stage	Action	Next Stage
1	Data Collection	Collect raw data from sources	Data Preparation
2	Data Preparation	Clean and format data	Model Training
3	Model Training	Train model on prepared data	Model Evaluation
4	Model Evaluation	Test model accuracy and performance	Model Deployment
5	Model Deployment	Deploy model to production	Monitoring & Maintenance
6	Monitoring & Maintenance	Track model performance and update	Data Collection (if needed)
7	Data Collection	Cycle repeats if model needs retraining	Data Preparation

💡 Cycle repeats to improve model or stops if model performs well

Status Tracker

Stage	Start	After Step 1	After Step 2	After Step 3	After Step 4	After Step 5	After Step 6	After Step 7
Stage	Data Collection	Data Preparation	Model Training	Model Evaluation	Model Deployment	Monitoring & Maintenance	Data Collection	Data Preparation

Key Moments - 3 Insights

Why does the lifecycle return to Data Collection after Monitoring?

Because monitoring may show the model needs new or updated data, so the cycle restarts to improve the model, as seen in steps 6 and 7 of the execution_table.

Is Model Deployment the last step in the ML lifecycle?

No, after deployment, the model must be monitored and maintained to ensure it works well over time, shown in step 5 moving to step 6.

What happens during Data Preparation?

Raw data is cleaned and formatted to be usable for training, which is the action in step 2 of the execution_table.

Visual Quiz - 3 Questions

Look at the execution_table, what is the stage after Model Training?

A Model Deployment

B Model Evaluation

C Data Preparation

D Monitoring & Maintenance

💡 Hint

Check row 3 in the execution_table under Next Stage column.

At which step does the lifecycle start monitoring the model?

A Step 6

B Step 5

C Step 4

D Step 7

💡 Hint

Look at the Current Stage column in step 6 of the execution_table.

If the model performs well, what happens to the cycle?

A It skips Monitoring & Maintenance

B It repeats from Data Collection

C It stops after Model Deployment

D It jumps directly to Model Training

💡 Hint

Refer to the exit_note explaining when the cycle stops.

Concept Snapshot

ML lifecycle stages:
1. Data Collection - gather raw data
2. Data Preparation - clean and format data
3. Model Training - build model
4. Model Evaluation - test model
5. Model Deployment - release model
6. Monitoring & Maintenance - track and update
Cycle repeats if needed to improve model

Full Transcript

The ML lifecycle starts with collecting data, then preparing it for use. Next, a model is trained on this data. After training, the model is evaluated to check its quality. If good, it is deployed to production. Then, the model is monitored to ensure it keeps working well. If monitoring shows problems or new data is available, the cycle restarts from data collection to improve the model. This cycle helps keep ML models accurate and useful over time.