Using human intelligence to train the machine
Ashley Rogers, Lead ML Engineer at a $22.3 billion American omni-channel retail giant, was setting up a model for comparing product listings across competitor sites covering 20K+ URLs and 100 product types. Recognising that it would be challenging to maintain accuracy with such a high volume ask, she looped in Infosys BPM for support. This case details how a team from Infosys BPM swiftly compared all product listings, annotated the variations, and labelled over 21,000 datasets with 100% accuracy, resulting in valuable competitor insights for the retailer and improved model performance.
Approach summary:
- Conducted comparative analysis across product listings
- Annotated differences in product attributes
- Integrated internal annotation tools with AI/ML pipelines
- Labelled 21,000+ datasets
- Ideated recommendations for data enrichment
Key benefits:
Accelerated model development via early completion 
High data quality levels achieved with zero rework 
AI/ML model accuracy enhanced 
