RAISE — Resume Analysis & Inference Serverless Engine
RAISE is a serverless cloud-native platform that automates the ingestion of candidate resumes across multiple formats (including PDF, DOCX, and DOC), utilizing advanced large language models to extract rich, structured metadata. This data is then persisted to a high-performance NoSQL database or published to a managed message queue for downstream processing.
Core Features
- Automated resume ingestion (PDF, DOCX, DOC) via SQS-buffered event-driven processing
- Structured data extraction via AWS Bedrock (Claude Sonnet 4.6)
- Extracts: name, email, phone, professional summary, work experience, education, skills, certifications
- Configurable output: DynamoDB, SQS, or both
- Built-in ingestion reliability: Dead Letter Queue with one-click redrive
- Multi-signal CloudWatch alarms with SNS email alerts
- CloudWatch Dashboard — single-pane view of all ingestion signals
- Infrastructure-as-code with AWS CDK TypeScript
Stack Components
| Layer | Service | Purpose |
|---|---|---|
| Ingestion | Amazon S3 | Receives resume uploads (PDF, DOCX, DOC) |
| Event Buffer | Amazon SQS — Ingestion Queue | Buffers S3 upload events; provides automatic retry via visibility timeout |
| Compute | AWS Lambda | Fetches document from S3, invokes Bedrock, routes structured output |
| Inference | AWS Bedrock (Claude Sonnet 4.6) | Extracts structured candidate data from resume |
| Storage | Amazon DynamoDB | Stores structured candidate profiles |
| Downstream | Amazon SQS — Output Queue | Publishes extracted profiles for downstream consumers |
| Reliability | Amazon SQS — Dead Letter Queue | Captures failed events immediately on first failure; one-click redrive to recover |
| Alerting | Amazon SNS | Delivers email alerts when any CloudWatch alarm fires |
| Observability | Amazon CloudWatch | 4 alarms covering DLQ depth, Lambda errors, duration, and queue age; unified dashboard |