02 /AI / DISTRIBUTED SYSTEMS
MICRORAG.
Upload a resume and a job post and get a report on what's missing and how to close the gap. Built to serve many people at once.
How it works
One pipeline. Four services. Scales by adding workers.
A RAG pipeline split into four services behind a Valkey queue, so slow PDF parsing never blocks the API and throughput scales by adding workers. A LangGraph orchestrator does the resume-versus-job reasoning, and every hop is traced in LangSmith so a bad answer can be followed back to the step that caused it.
The hard parts
- Splitting ingestion, embedding and retrieval so each can scale on its own.
- Keeping long-running jobs asynchronous behind a queue.
- Tracing a request across several services.
Results
- Deployed on AWS EC2 behind an Application Load Balancer through GitHub Actions.
- Scales horizontally by adding workers or instances.
Stack
Reach the end of the file to complete this operation
