NashTech

The art of human-AI collaboration: A case study in model improvement

The art of human-AI collaboration: A case study in model improvement

Introduction

For over 16 years, NashTech has been a trusted partner, providing data management solutions that have fuelled the exponential growth of our client’s online shopping platform. Our comprehensive data qualification process ensures product information accuracy exceeding 99%, surpassing the client’s initial target of 98.5%. This meticulous approach has directly contributed to attracting millions of satisfied shoppers and generating significant revenue for the client across diverse international markets.

Company overview

Our client is one of the world’s largest multinational technology companies specialising in online services and products, including advertising, search, and cloud computing. One of the services that the client offers to its users is the ability to search for products online comparing prices between different vendors.

The challenge

Our client possesses a legacy portfolio of AI and machine learning systems. These systems have generated a significant volume of unstructured data, necessitating data quality initiatives to ensure its accuracy and optimise its utility. 

A comprehensive data integrity audit is required to validate online product listings on the shopping platform, ensuring the accuracy of product information presented to consumers. 

Challenge – To protect their brand and increase revenue, our client needed a way to ensure that the product information they manage and is accessed by customers was up-to-date, complete and accurate. 

To guarantee the reliability and quality of content produced by AI and ML systems, human-in-the-loop remains essential. This oversight allows identification and rectification of potential inaccuracies, ultimately refining the final output. 

The project demanded a rigorous 98% accuracy threshold across 28 languages. 

The approach and solution

To optimise the accuracy of our client’s automation and machine learning models, NashTech implemented a comprehensive data enhancement program, leveraging manual data checks, qualification and editing processes to ensure high-quality data that fuels superior performance. 

Key services: 

  • Multilingual data quality audit: Our team of qualified linguists meticulously validates displayed product information across 28 languages, including elements like name, category, variant, price, currency and availability. This ensures the accuracy and consistency of data presented to the client’s customers. 
  • Image quality and relevance evaluation: NashTech conducts a thorough audit of images processed by the client’s algorithms. This evaluation assesses image quality, extraction accuracy and, most importantly, the image’s relevance to the corresponding product offering. 
  • Virtual try-on AI optimisation: NashTech audits, annotates, detects and qualifies the AI-generated images in the client’s dataset. We then enhance the AI logic to produce an updated dataset with improved image quality. This process helps train and refine the AI models’ performance in generating high-quality virtual try-on images.  
  • AI data labelling and model training: NashTech helps client’s AI in complex data labelling tasks for end-user purchased product reviews, detecting fake reviews, completing the dataset and reflecting the results to enrich large language model (LLM). We train AI models to automatically label complex cases, thus, improving their accuracy. 

Furthermore, our team proactively identifies and reports any issues or inconsistencies within the client’s platform, ensuring a seamless user experience. 

Results and benefits

For over 16 years, NashTech has been a trusted partner, providing data management solutions that have fuelled the exponential growth of our client’s online shopping platform. Our comprehensive data qualification process ensures product information accuracy exceeding 99%, surpassing the client’s initial target of 98.5%. This meticulous approach has directly contributed to attracting millions of satisfied shoppers and generating significant revenue for the client across diverse international markets. 

Measurable success: 

  • We consistently deliver a product data accuracy rate exceeding 99%, exceeding the client’s initial target of 98.5%.
  • Our expertise enables the efficient auditing of over 100 million data points annually, ensuring a seamless customer experience.
  • We provide comprehensive data management services in 28 languages, catering to the client’s global audience.
  • By continuously identifying and rectifying inaccuracies, we’ve helped refine the client’s AI/ML systems, enabling them to become more scalable and efficient. 

Read more case studies

The art of human-AI collaboration: A case study in model improvement

For over 16 years, NashTech has been a trusted partner, providing data management solutions that have fuelled the exponential growth of our client’s online shopping platform. The approach has...

From rising above adversity to riding the wave of digital transformation in the education sector

Explore how NashTech help Trinity College London ride the wave of digital transformation in the education sector

AWS
AWS
Migrating and modernising the virtual learning environment to AWS for an enhanced experience

The migrated and modernised Moodle infrastructure means that The Open University can now take advantage of cloud benefits.

AWS
AWS

Our partnerships

Scroll to Top
SAMPLE TITLE
sample short
sample heading lorem isump
FREE WHITEPAPER
Unlock the power of knowledge with our new whitepaper
“Elevating User Experience for Product Owners”
FREE WHITEPAPER
Unlock the power of knowledge with our new whitepaper
“Elevating User Experience for Product Owners”