Building custom tools in Agentic AI - Model Pipeline Trace

Model Pipeline - Building custom tools

This pipeline shows how to build a custom AI tool that learns from data, improves over time, and makes predictions to help with specific tasks.

Data Flow - 6 Stages

1Data Collection

1000 rows x 6 columns→Gather raw data relevant to the custom tool's task→1000 rows x 6 columns

User activity logs with 6 features like time, action type, and context

↓

2Data Preprocessing

1000 rows x 6 columns→Clean data, handle missing values, and normalize features→1000 rows x 6 columns

Missing times filled, action types encoded as numbers, values scaled between 0 and 1

↓

3Feature Engineering

1000 rows x 6 columns→Create new features to better represent the data→1000 rows x 8 columns

Added features like time of day category and action frequency count

↓

4Model Training

800 rows x 8 columns→Train model on 80% of data to learn patterns→Trained model

Model learns to predict user intent from features

↓

5Model Evaluation

200 rows x 8 columns→Test model on 20% unseen data to check accuracy→Accuracy score and loss value

Model achieves 85% accuracy on test data

↓

6Prediction

1 row x 8 columns→Use trained model to predict output for new input→Prediction result

Model predicts user wants to schedule a meeting

Training Trace - Epoch by Epoch

Loss
0.7 |****
0.6 |*** 
0.5 |**  
0.4 |*   
0.3 |*   
     1 2 3 4 5 Epochs

Epoch	Loss ↓	Accuracy ↑	Observation
1	0.65	0.60	Model starts learning basic patterns
2	0.50	0.72	Loss decreases, accuracy improves
3	0.40	0.80	Model captures more complex relationships
4	0.35	0.83	Training stabilizes with good accuracy
5	0.30	0.85	Final epoch shows best performance

Prediction Trace - 4 Layers

Layer 1: Input Layer

Layer 2: Hidden Layer with ReLU

Layer 3: Output Layer with Softmax

Layer 4: Prediction

Model Quiz - 3 Questions

What happens to the data shape after feature engineering?

A It decreases from 8 to 6 columns

B It stays the same at 6 columns

C It increases from 6 to 8 columns

D It changes from rows to columns

Key Insight

Building custom tools involves preparing data carefully, training a model that improves over time, and using the model's predictions to assist with specific tasks effectively.