ExploreTech PRO
A Python microservice behind a conversational product: crawling, embeddings and a graph database feeding an agent that answers over a WebSocket, with a Laravel dashboard alongside it.
- 3
- services in one product
- S3
- offsite backups I provisioned
- Python
- Graph DB
- Embeddings
- Crawling
- WebSocket
- AWS
- Docker
The problem
The product needed to answer open questions about a large, constantly changing body of vendor and city data — the kind of thing a search box handles badly and a plain chatbot gets confidently wrong.
My role
Cross-functional: the Python service is home base, but the work spans backend, frontend, QA and release management as needed.
What I built
- 01
A microservice split along hard boundaries — retrieval and graph code never imports the management side, and crawlers and schedulers are never exposed as endpoints.
- 02
Crawling and embedding pipelines keeping the graph current, with data clean-up passes for the messy source material.
- 03
An agent serving answers over WebSocket to a separately hosted chat frontend.
- 04
Set up the AWS account and offsite S3 backups for the platform's data.
Outcome
Ships on a staging-to-production flow with release notes and rollback snapshots kept per deploy.
I work across every part of it rather than one repo, which is how the awkward integration bugs get found.