The PeripherAi teams’ goal was to develop a sales enablement tool for start-ups and SME’s, a tool that would support small companies and SMEs to optimise their sales process by using their data to optimise and expedite their sales effort.

PeripherAi wanted to gather these data sources to create an engine that allows them to find the optimal sales process in the most effective manner, then accelerate the process further using machine learning.

The data the PeripherAi team were looking at utilising would be from the CRM, analysing communications with clients such as email, voice and/or video call recordings.

PeripherAi needed the help of a partner that had both machine learning expertise with AWS services, as well as a deep knowledge of data ingestion and workflows.

Firemind’s team have a capable understanding of ML as well as building data workflows into ML initiatives.

Of the many critical stages in building ML models, data ingestion plays a significant role in the result of data-driven initiatives for organisations. Since data comes from different sources, building a robust data ingestion pipeline is essential to feed the desired and relevant data into ML services, generating high quality ML Models to better improve the engagement with prospects and customers.

The ingestion phase was designed for the ingestion and categorisation of various document types prior to machine learning techniques being applied. This will allow for the baseline deployment to be compatible with other ML initiatives such as the capability to use Amazon Translate, Amazon Textract and Amazon Transcribe, as the product set starts to evolve.

As advised, this design focuses on the initial data ingestion module. The data ingestion solution will be comprised of the following:

• Audio transcripts in JSON format are uploaded to the NLP/Raw S3 bucket.
• The document registrar Lambda function is invoked. This function calls the metadata services API to register the document and in turn receives a unique ID. This unique ID is added to the document as a tag, and the metadata is registered within the DynamoDB table; document registry.
• After the document metadata is registered with the metadata services, the DynamoDB document registration stream is invoked to start the document classification Lambda function, which exams the metadata registered on the document. The result of this examination is written back to the metadata services.
• The metadata registration of the previous step invokes a DynamoDB Pipeline Operations Stream, which invokes the Document Extension Detector Lambda function.
• This function examines the incoming file formats and separates the files into separate buckets for processing using the various services that are available on AWS.

Rather than manual or customer filled run books, this new solution provides fully automated data capture, savings the PeripherAi team significant time and resource as they no longer need to manually edit and adjust field content.

Utilising Amazon Comprehend and Amazon QuickSight will ensure trained data models are made through analysis of voice recordings.