PREDICTif Meeting

Discussed AI model conversation about NIGP commodity codes

Reviewed companies doing similar classification work

Examined NIGP code structure and descriptions

Explored sample vendor data and contract information

Shared screen to orient about the project

Discussed AI model conversation about NIGP commodity codes

Mentioned companies doing similar classification work

Explained NIGP code structure and its use in procurement

Detailed the 11-digit NIGP code structure

Explained class, item, and group divisions within the code

Mentioned the number of codes at different digit levels

Reviewed sample vendor data for VSN Sports

Discussed awarded items and associated NIGP codes

Explored vendor websites and PDF catalogs as data sources

Examined manually created spreadsheets with NIGP codes

Introduced the proposal for a custom AWS solution

Discussed the benefits of building vs. buying a solution

Explained the role of Predictif Solutions as an AWS partner

Addressed questions about AWS services and infrastructure

Introduced the proposal for a custom AWS solution

Explained the focus on creating a scalable, low-cost solution

Discussed the benefits of building vs. buying a solution

Addressed the proof of concept approach

Discussed AWS infrastructure and availability zones

Addressed questions about data center locations

Explained the flexibility in choosing deployment locations

Discussed integration with existing OVH environment

Discussed the current manual process for NIGP code assignment

Explored the SQL database structure and tables

Identified the main problem with NIGP code assignment

Introduced the concept of using AI for code classification

Discussed the time-consuming nature of manual NIGP code assignment

Explored examples of manually created spreadsheets

Identified the need for automation in the process

Introduced the potential for AI to assist in classification

Explored the SQL database structure and tables

Discussed the relationship between contracts, items, and NIGP codes

Identified the main problem with NIGP code assignment in the database

Introduced the need for an automated solution

Introduced the technical approach using AWS services

Explained the use of Amazon Bedrock for foundational models

Discussed the RAG (Retrieval Augmented Generation) approach

Addressed questions about model selection and evaluation

Introduced the use of AWS services for the solution

Explained the role of Amazon Bedrock for hosting foundational models

Discussed the benefits of using pre-trained models

Addressed questions about model selection and evaluation

Explained the RAG (Retrieval Augmented Generation) approach

Discussed the use of vector databases for efficient searching

Addressed questions about similarity search and semantic scoring

Compared RAG to fine-tuning approaches

Presented the AWS architecture diagram

Explained the data ingestion and processing pipeline

Discussed the classification model and confidence scoring

Addressed questions about workflow management and step functions

Presented the AWS architecture diagram

Explained the data ingestion process from external sources

Discussed the use of Amazon S3 for data storage

Introduced the ETL process using Amazon Glue

Explained the classification model using Amazon Bedrock

Discussed the confidence scoring mechanism

Introduced the verification workflow for human review

Addressed questions about API endpoints and integration

Discussed the deployment options and infrastructure considerations

Introduced the project team and roles

Explained the project timeline and milestones

Addressed questions about knowledge transfer and documentation

Discussed the deployment options within AWS

Addressed questions about data privacy and security

Explained the use of virtual private clouds (VPCs)

Discussed the integration with existing OVH infrastructure

Introduced the project manager and cloud solutions architect roles

Explained the project timeline and duration

Discussed the knowledge transfer and documentation process

Addressed questions about project milestones and updates

Discussed the business drivers for the project

Identified technical gaps in the current process

Explained how the proposed solution addresses these gaps

Addressed questions about ROI and employee impact

Discussed the main business drivers for the project

Explained how the solution addresses manual process inefficiencies

Discussed the potential for faster vendor onboarding

Addressed questions about return on investment (ROI)

Identified technical gaps in the current manual process

Explained how the AI-driven solution addresses these gaps

Discussed the benefits of automating NIGP code assignment

Addressed questions about employee impact and job roles

Discussed the process for evaluating different foundational models

Explained the reporting and recommendation process

Addressed final questions about project scope and next steps

Concluded the meeting with action items and follow-up plans

Explained the process for evaluating different foundational models

Discussed the criteria for model selection (performance, accuracy, cost)

Addressed questions about model updates and future improvements

Explained the reporting and recommendation process

Addressed final questions about project scope and deliverables

Discussed the process for sending additional product data

Explained the next steps for proposal review and approval

Concluded the meeting with action items and follow-up plans