Slide 1: Intelligent strategies for tomorrow's testing
Webinar Intelligent strategies for tomorrow’s testing © 2024, Nagarro
Slide 2: Agenda
Agenda Lightning Talks 1 Testing of AI-based applications Vimmi Walia 2 Self-healing pipelines Manisha Mittal 3 Accessibility Testing Anamika Mukhopadhyay
Slide 3: Testing of AI-based applications
Intelligent strategies for tomorrow‘s testing Testing of AI-based applications Vimmi Walia
Slide 4: An insight into the AI world
An insight into the AI world Diffusion GANs Variational Structure Transformers Autoencoders Image Discovery Classification Customer Meaningful Fraud Retention Compression Feature Generative AI Elicitation detection Dimensionally Classification Big Data Reduction Diagnostics Visualisation UNSUPERVISE
Slide 5: Traditional apps vs AI apps
Traditional apps vs AI apps Output System of rules Provide input Traditional apps Define UI design Development requirement Agile SDLC Define rules Release UAT QA USERS Tunes algorithm DEVELOPERS Problem & Goal Selects data Machine Learning SDLC Definition AI apps Learning Data Delivery & Data acquis
Slide 6: AI SDLC – Zoom-in
AI SDLC – Zoom-in Business user Data & ML QA Data scientist Objectives Problem definition Stakeholders & Data sources comm. Resources Feasibility constrains assessment Existing practices Milestones Data steward/ Data ML solution Problem & goal provider acquisition definition Data cleansing Documenta
Slide 7: AI is not black box
01 AI is not black box Algorithms : AI is a collection of complex mathematical functions/algorithms like Linear regression, Decision trees, Naive Bayes. 02 01 Probability: AI applications are based on probabilistic logic, hence the results are not black Data collection and white. We talk in term deg
Slide 8: Key takeaways
Key takeaways Customized test Lack of determinism Data quality Bias Model and data drift Explainability approach for every use Exhibit non- Ensuring the quality, Source Data may be The data that the model Many AI models, case deterministic behavior. relevance, and biased, and the data was trained on
Slide 9: Automated anomalies detection with self-healing pipelines
Intelligent strategies for tomorrow‘s testing Automated anomalies detection with self-healing pipelines Manisha Mittal
Slide 10: Problem statement
Problem statement Are you facing these challenges in CI/CD & scared of losses which can occur due to OUTAGES ? • Manual intervention for error detection and resolution • Manual inspection of data & logs is unmanageable & time consuming • Limited scalability with complex systems • Increased downtime
Slide 11: Self-healing pipeline model
Self-healing model aims at an innovative approach for addressing the challenges faced by traditional deployment techniques. These pipelines are designed to autonomously identify and resolve issues without requiring manual intervention. Self-healing By leveraging automation and intelligent algorithms
Slide 12: Key features of the DevOps self-healing model
01 Anomaly detection Monitor ->Detect ->Trigger Key features Structured pipeline logs are used as input to AI-based anomaly detection algorithms to identify unusual or unexpected patterns in pipeline executions like sudden failures. For example, AI model can detect the of the DevOps failure cause of
Slide 13: How it works
How it works Real-time data analysis before Self Heal AI Model production Known anomaly type Production deployment Data/Log Data Monitor Anomaly collection preprocessing metrics detection Deployment Monitoring Unknown anomaly type Trigger alerts Self-heal Trigger alerts Capture Manual feedback & ins
Slide 14: Self-healing pipeline scenarios
Network or connectivity issues Pipeline execution fail if there are network issues or connectivity problems between CI/CD tools and external resources, such as source code repositories, artifact repositories, or external services. Insufficient permissions If the CI/CD tools user or service account d
Slide 15: Challenges in achieving the self-healing
Challenges in achieving the self-healing Cultural shift Security concerns Algorithm selection Accurate decision Monitoring and Adopting a self-healing Automated responses raise Selecting appropriate Ensuring precise anomaly feedback loops mindset necessitates a security concerns, algorithms for anom
Slide 16: Key takeaways
Key takeaways Autonomous problem solving: Detects and corrects issues without human intervention Resilience and reliability: Reduces downtime, ensures smooth operations Cost-efficiency: Saves time, resources, and mitigates risks Scalability: Adapts to changing demands seamlessly Continuous improveme
Slide 17: Accessibility Testing
Intelligent strategies for tomorrow‘s testing Accessibility Testing Anamika Mukhopadhyay
Slide 18: What is Accessibility?
What is Accessibility? • Refers to the design and development of products, services, and environments that can be accessed and used by individuals with disabilities • It aims to eliminate barriers and provide equal access and opportunities for everyone, regardless of their abilities No longer a choi
Slide 19: Tomorrow's testing need
Tomorrow’s testing need An estimated 1.3 billion people experience The accessibility market is expected to grow significant disability. This represents 16% of at a compound annual growth rate (CAGR) of the world’s population, or 1 in 6 of us (WHO). 6.20 % from 2023 to 2028. In 2020, companies faced
Slide 20: Why is Accessibility needed?
Why is Accessibility needed ? User experience Legal compliance Equal access Social responsibility Business benefits Innovation and creativity The power of the web is in its universality. Access by everyone regardless of disability is an essential aspect. -Tim Berners-Lee, Director of the W3C and inv
Slide 21: Some consequences of lack of a11y
Some consequences of lack of a11y 2005 2012 vs vs 2017 2017 2018 https://www.whoisaccessible.com/guidelines/largest-web-accessibility-lawsuits/ Dennis Maria Himelda vs vs vs Haynes Mendizabal Mendez 2019 2022 Guillermo Rafael vs vs Robles Cordero Reference: • https://www.whoisaccessible.com/guidelin
Slide 22: Laws/Regulations
Laws/Regulations United States United Kingdom Americans with Disabilities Act Disability (ADA) Discrimination Act (DDA) European Union Canada Australia European Accessibility Act Accessible Canada Act Disability Discrimination (EAA) (ACA) Act (DDA)
Slide 23: Testing for success: A comprehensive accessibility strategy
Testing for success: A comprehensive accessibility strategy Manual testing Automated testing Making use of assistive technology like screen Integrate plug-ins like Axe into the development readers, speech recognition software and workflow to expedite testing process special adaptive keyboards Browse
Slide 24: Tools and Domains
Tools and Domains Automated Testing Tools Domains Web BFSI Telecom Accessibility Insights Wave Axe Tenon pa11y Lighthouse Travel and Logistics Browser Extensions Mobile Aviation Automotive Ecommerce Accessibility Insights Wave Axe Accessibility Scanne Axe Dev Tools Energy and Utilities Entertainment
Slide 25: By 2025, the number of differently abled people employed will triple
By 2025, the number of differently abled people employed will triple, reducing barriers to access Let's work together to make a positive impact on our customers and our community.
Slide 26: Thank you
Thank you aqt@nagarro.com © 2024, Nagarro