Understanding AI Fundamentals for Quality Engineering
Deepening My Fundamentals: Data, Models, and How AI Really Works
Today is Day 2 of my AI learning journey, and I focused on strengthening my foundational understanding of how AI works behind the scenes. These concepts are essential because they show why AI behaves the way it does — and how QA engineers like me can apply AI meaningfully in testing.
Here are the key topics I explored today:
- Machine learning fundamentals - Labeled vs. unlabeled data - Structured vs. unstructured data - Inferencing - Deep learning fundamentals & neural networks - Natural Language Processing (NLP) - Generative AI fundamentals - Foundation models & their lifecycle
1. Machine Learning Fundamentals
Machine Learning (ML) is the science of teaching computers to learn patterns from data instead of explicitly programming every rule.
In QA, ML shows up in:
self‑healing test locators
predictive defect analysis
anomaly detection
intelligent test suggestions
At its core:
- Data goes in → Model learns patterns → Model predicts outcomes
2. Labeled vs. Unlabeled Data
Data is the fuel of all AI models, but not all data is the same.
Labeled Data
This data has predefined tags or answers. Example:
Email + tag “Spam”
Image + tag “Cat”
Useful for:
Classification
Regression
Unlabeled Data
No tags, no predefined answers. Example:
Raw text
Screenshots
Logs
Useful for:
Clustering
Pattern discovery
Anomaly detection
QA relevance:
Labeled data powers supervised learning (e.g., classifying defects)
Unlabeled data helps uncover patterns in large test logs
3. Structured vs. Unstructured
Structured Data
Organized in tables, rows, and columns.
Examples:
Test execution logs
Pass/fail metrics
Bug counts
Unstructured Data
Messy, free‑form data without a fixed structure.
Examples:
Screenshots
Video recordings
Free‑text defect descriptions
Chat messages
Most testing artifacts are unstructured, which is why AI is critical — it can interpret meaning where traditional systems cannot.
4. Inferencing
Inferencing is the step where a trained AI model uses what it has learned to make predictions.
Simple way to think about it:
Training = learning
Inferencing = applying the learning
In QA:
AI suggests test cases → inferencing
AI detects anomalies in logs → inferencing
AI predicts flaky tests → inferencing
This is where AI becomes useful.
5. Deep Learning & Neural Networks
Deep learning uses multi-layered neural networks to process complex data like:
images
videos
audio
natural language
Neural networks work like layers of “digital neurons” passing information forward.
Why QA teams care:
visual validation
log summarization
test failure classification
video-based analysis of user flows
Deep learning enables more “human-like” reasoning in testing tools.
6. Natural Language Processing (NLP)
NLP allows machines to understand and generate human language.
This is the technology behind:
AI-written test cases
Requirement summarization
Defect analysis and categorization
Converting plain English to Playwright/Selenium test scripts
Chatbots used for test data generation
NLP is the backbone of AI-augmented QA.
7. Generative AI Fundamentals
Generative AI creates new content based on patterns it has learned.
Examples:
writing test scripts
creating test data
generating bug descriptions
converting Selenium → Playwright
writing documentation
summarizing logs
GenAI is built on large “foundation models.”
8. Foundation Models & Their Lifecycle
Foundation models are massive pretrained models that learn from huge amounts of data and can be adapted (fine‑tuned) for thousands of tasks.
Lifecycle of a Foundation Model:
Pre-training
Learning general patterns from huge datasets.Fine-tuning
Adapting the model for specific domains (e.g., testing).Deployment
Using the model in real workflows.Monitoring
Ensuring outputs remain correct.Retraining
Improving the model with new data.
Why this matters for QA:
These models can understand your application, predict issues, and even write tests.
My Day 2 Takeaways
AI is built on simple ideas that scale with massive data.
Understanding the fundamentals helps me see where AI fits in QA.
Most QA data is unstructured — perfect for AI.
Foundation models are the engines behind Generative AI.
These concepts will help me use AI better in Playwright automation.
Day 2 Sign-Off
Another day, another set of insights.
AI is starting to make more sense — not as something magical, but as a series of smart, logical systems that can amplify our abilities as testers.
See you on Day 3.
— Hema
