AI Assistants in Healthcare: An Open Source Starter Pack for Developers

Hey Rasa Community,

We just open sourced Medicare Locator - an example AI assistant to help developers in healthcare with building their own:

Would love to hear your thoughts on it and also any cool use cases you’ve seen for AI assistants in healthcare!


Hey ! I checked your repo to see which pipeline/policy you are using for this bot, and I’m wondering why you do not have a fallback policy in your core_config ? :thinking:

Hi Alex,

Looks awesome. I went through the tutorial and something that strike me is it seems that now you are enabling verticals using the rasa-stack where i see you are using the spacy en language model. Any chance that those embeddings are fine tuned to address the needs of a particular vertical domain?

Is it something you consider as a strategy towards other vertical domain so to release more fine-tuned custom embeddings or nlu classifiers ?

It could be interesting to release some vertical dialogues addressing the most common use-cases in the industry that can be then finetuned to a particular business specially from the perspective of rasa core because then most developers can already start off with something quick and continue working with it

This is really interesting. We are actually working on something similar at our company. Helping customers to locate hospitals, pharmacies etc, provide users with definition of basic medical terms.

1 Like

Sorry for my super late reply here everyone!

@huberrom Thanks! There is no special reason for not adding it; we’ve mostly been focussed on the connection to the Medicare APIs. Would be a great addition to it and we’d love to see a contribution if you want to support the project: GitHub - RasaHQ/medicare_locator: 🏥Medicare Locator - Open source starter pack for developers to build contextual chatbots and AI assistants in healthcare

@souvikg10 Thanks for your message! In this version, there are not special vertically-specific language models but we see companies more and more creating their own based on specific use cases. We’d love to offer more as well but that’s a question of resources on our end. :slight_smile: What do you think would be another useful starter pack to create?

If you are creating a financial industry starter pack, i would love to help. I think starter pack is a great idea to engage with verticals. We are also trying internally to build a starter pack on Rasa to showcase and engage with the business departments regarding the possibilities.


Thanks @souvikg10 - will keep that in mind and get in touch if we do one for financial services industry!

Hi @alexweidauer, I have gone through step by step implementation of Medicare locator now. I have successfully implemented this Medicare locator in 2019 end. But now its not working. I am following below Steps:- 1)- creation of new environment using anaconda prompt. 2)- installation of Python conda install python=3.6. 3)- i have cloned your git code locally in my system. 4)- pointing to the local directory and then running pip install -r requirements.txt 5)- I am running make commands directly like for make train-core i am running python3 -m rasa_core.train -d domain.yml -s data/core -o models/current/dialogue -c core_config.yml on running this i am getting this error AttributeError: module ‘tensorflow’ has no attribute ‘Session’. I searched and applied several solutions but it is not getting fixed, Request you to guide and share if you have any updated code for the same which is working. Thanks in advance for your support.