Case Study
Reklaim
AI-Powered Product Identification Using Amazon Bedrock
June 2025 | LOCATION
INDUSTRY
Retail
Project Type
GenAI Platform Migration
BaSIS Services
Generative AI, AI/ML
Technology Leveraged
Amazon Bedrock, Amazon SageMaker, Amazon EC2, Amazon Lookout For Vision, Amazon EC2, Amazon EKS, Amazon SQS, Amazon API Gateway, Cloudwatch
Reklaim
AI-Powered Product Identification Using Amazon Bedrock
Problem / NEED
Reklaim, a small to medium sized business (SMB) in the luxury resale industry, needed a highly accurate, scalable solution for identifying high-end handbags and watches to remain competitive in the fast-growing luxury resale market. Their original platform, powered by OpenAI models, struggled with classification accuracy (only 70–80%), lacked real-time performance, and could not scale efficiently during peak auction data loads.
The low accuracy rates from OpenAI models prevented Reklaim from setting competitive prices, maintaining customer trust, and scaling operations. As a result, Reklaim had to undergo labor-intensive manual verification processes, which was unsustainable and hindered growth. Their legacy infrastructure had also become a barrier to scaling efficiently.
Solution
Basis partnered with Reklaim to replatform their identification system onto a fully AWS-native, GenAI-powered architecture using Amazon Bedrock and SageMaker. Rather than a lift-and-shift migration, the solution was re-architected to take full advantage of multimodal GenAI capabilities.
Key components included:
- Multimodal AI Processing with Amazon Bedrock to analyze product imagery, metadata, and certificates simultaneously for improved classification accuracy.
- Custom-trained models for tasks like color classification, model number prediction, and brand identification using YOLOv8, CLIP, and encoder-decoder architectures.
- Advanced OCR pipelines for extracting serial/model numbers from product documents and dials using PaddleOCR, LayoutLM, and Meta’s Segment Anything Model.
- Scalable infrastructure using Amazon EKS and EC2 with auto-scaling, SQS queues, and API Gateway to support real-time auction data flows.
- Rigorous validation via Amazon SageMaker, with phased pipeline cutovers and full rollback protection.
The architecture enabled end-to-end automation across product ingestion, image analysis, condition assessment, and brand/model verification.
RESULTS
Reklaim’s post-migration platform achieved a 95% accuracy rate, up from 70–80%, representing a 25-point gain in classification performance. This directly improved pricing precision and customer trust on their platform.
Other key outcomes:
- 99.95% system uptime with sub-second response times, enabling real-time identification workflows.
- Expanded coverage across 9 major luxury brands and hundreds of bag and watch models.
- Cost savings and scalability from AWS-native infrastructure and managed GenAI services.
- Reduced operational overhead, eliminating the need for manual validation at scale.
Reklaim’s AWS-based GenAI platform now positions the company as a technology leader in the luxury resale market, with a competitive edge built on accuracy, speed, and trust.
Technology Leveraged
