Tailoy and school lists
Tailoy was looking to optimize its process of loading, processing and sending school lists through a scalable, automated and artificial intelligence-based solution. With the help of BigCheese, a PoC was implemented using AWS services, validating the technical feasibility of the project.
The challenge
The main challenge was to develop an automated solution that would allow customers to upload school supply lists in various formats (PDF, images, etc.), process them to extract the relevant items and return a list of products available in Tailoy’s e-commerce.
The system was required to be fast, accurate and able to integrate with Tailoy’s platform without affecting the user experience.
The solution
BigCheese designed and implemented a serverless architecture based on AWS, using artificial intelligence to extract information from school rosters and improve the efficiency of the process. The solution included:
- List loading:
An event-driven process was triggered when a user uploaded a list to Amazon S3.
Amazon Bedrock (Sonnet 3.5 v2) was invoked to extract the relevant information.
Data was stored in Amazon DynamoDB in JSON format.
2. List processing:
AWS Lambda was used to process the data and find the products in the Tailoy catalog.
An AI-based search engine was invoked that mapped the items to their SKU codes in the database.
3. Sending lists:
A Lambda inside a VPC was connected to Amazon RDS for Magento, sending the processed data.
A timeout of 29 seconds was defined due to API Gateway limitations.
The implemented solution was fully serverless, scalable and cost optimized.
AWS applied in this challenge
Amazon S3: Storage of lists uploaded by users.
Amazon Bedrock: AI processing for data mining.
Amazon DynamoDB: NoSQL database to store the processed lists.
AWS Lambda: Serverless functions for processing and sending data.
Amazon API Gateway: Exposure of APIs for customer interaction.
Amazon RDS: Relational database integrated with Magento.
Amazon SQS: Asynchronous event management for efficient processing.
The results
90%
During the proof of concept, BigCheese’s solution achieved 90% accuracy in identifying items on school rosters, using Amazon Bedrock (Sonnet 3.5 v2) and AWS Lambda to structure and store the data in Amazon DynamoDB.
100%
100% successful integration with Tailoy’s search service was fully successful, enabling efficient and accurate execution. The system optimized product identification, improving the user experience.
100%
100% satisfaction within the expected times.
Response times were within the defined parameters, ensuring a smooth execution and an optimal user experience without interruptions or delays.
Tailoy’s PoC was 100% successful, validating that BigCheese’s proposed solution was viable and could be implemented in production. Now, Tailoy is evaluating the next phase to scale the solution and bring school roster automation to its entire digital ecosystem.
