Chatbot platform for a healthcare provider
- Industry
- Healthcare
- Type of work
- Chatbot architecture & MLOps
- Period
- 2018
- With
- ML Soluciones Informáticas (founding partner)
Confidential project: code and data are not public.
Architecture
- Channels WebChat ReactAPI RESTBot Connector
- Bot Bot Builder SDKshared libraryper-bot plugins
- Intelligence DialogFlow: intents and entitiesticketsQ&A · Elasticsearch
- Operations Google Sheets for agentsAirflow: retrainingfallback review
- Delivery Jenkinsdev → QA → UAT → prodVSTS
- Infrastructure AzureKubernetesDocker
Goal
Let an IT integrator build and sell chatbots for a clinic and other clients without rebuilding each one from scratch, and keep them trained without relying on developers.
What I did
- Designed the chatbot architecture: Bot Framework for channels, DialogFlow for language understanding, and a shared library with per-bot plugins.
- Designed the cognitive modules for tickets and for Q&A on Elasticsearch.
- Designed the continuous delivery pipeline with Jenkins, promoting each bot through dev, QA, UAT and production on Kubernetes on Azure.
- Designed the recurring operations: Airflow retrains and loads new intents from a spreadsheet that agents edit.