Finding the best candidate for every role, in seconds

A SaaS B2B business working with Cloud Combinator on AWS. The client is anonymised at their request.

The challenge had four focus areas, each of which shaped the design of the system.

At the centre of the solution is an Amazon Bedrock Knowledge Base, built on an Amazon OpenSearch Serverless vector store, that holds each client's job specifications and supporting context. When documents are uploaded to Amazon S3 they are tagged with client and requirement metadata, and an AWS Lambda function keeps the knowledge base in sync automatically.

The proof of concept was built against a clear success definition: a working Bedrock-powered workflow that evaluates and ranks CVs against a chosen job specification, keeps each client's data segregated, syncs new material automatically, and secures access, proving that the AWS environment is the right foundation for the client to take the service forward.

• Precise, grounded matching — An Amazon Bedrock Knowledge Base trained on the client's own hiring data, with metadata filtering by client and requirement, so results reflect the real role rather than a generic model.
• Ranked results in seconds — Each CV is scored from one to ten and sorted by likelihood of a positive outcome, with assessed CVs automatically filed by score into success and review paths in S3.
• Secure and safe by design — Amazon Cognito authentication, a VPC with private subnets, and Amazon Bedrock Guardrails to protect against bias and malicious input.
• Resilient data — Amazon S3 cross-region replication from the client's site to Ireland, and a DynamoDB store with snapshot backups.
• Handover — A working demonstration, the required materials and documentation were provided so the client's team could see and validate the service end to end.

The proof of concept was deliberately scoped to prove the pattern rather than ship a finished product. The clear path from here is a production-grade interface built for higher load, an

Optional hallucination-check step, and expansion into further AWS regions and models as the client grows the service into new markets.