PeripherAi helps start-ups and SMEs build scalable, efficient sales functions by integrating human knowledge with powerful data insights. Their goal is to enable small businesses to optimise sales processes and accelerate growth.
Challenge
PeripherAi aimed to create a sales enablement tool for start-ups and SMEs that could leverage CRM data, client communications, and machine learning to streamline and enhance sales performance. They wanted to automate the ingestion and categorisation of multiple data types – emails, call recordings, and transcripts -while enabling advanced analytics and sentiment analysis.
To achieve this, they needed a robust, scalable ingestion pipeline capable of integrating with future ML initiatives, reducing manual work, and producing reliable, high-quality training data for machine learning models.
Solution
Firemind built a data ingestion module designed to handle diverse data sources and feed them into ML-powered analytics. Audio transcripts in JSON format are uploaded to an Amazon S3 bucket, triggering a Lambda function to register documents in DynamoDB with unique IDs and metadata. This metadata is then processed through classification and extension detection Lambda functions, automatically sorting files for downstream ML services.
The system integrates Amazon Comprehend for sentiment analysis and is future-proofed for additional AWS services like Amazon Translate, Amazon Textract, and Amazon Transcribe. This automated approach replaces manual data preparation, saving significant time and enabling PeripherAi to focus on building advanced analytics and insights into their sales platform.
Services used
- Amazon Comprehend
- AWS Lambda
- Amazon DynamoDB
- Amazon S3
The Results
- 60% improvement in sentiment analysis accuracy
- Automated, scalable data ingestion pipeline
- Automated, scalable data ingestion pipeline Significant time and resource savings by removing manual preparation
- Future-ready architecture for advanced ML services
- Enhanced ability to analyse CRM and client communications data
