Architecture
.png)
What you may need to know:
<aside>
👉 This page is from the prior DS developers to the next developers on some points we think would be good for you to know ahead of time. look at the ideas board on Labs 20 Trello too.
</aside>
Trello
Lambda-School-Labs/Music-Meteorologist-ds
Import this into Postman
Postman | The Collaboration Platform for API Development
Ideas for DS Release Canvases
Code Optimization
- Review predict.py Sound_Drip object responsible for production application and optimize code for performance and legibility
Slider Functionality
- Integrate/ improve slider functionality
- Front end feature not implemented. Essentially allows users to generate playlist by sending the DS endpoint a JSON object containing the necessary attributes used for the model prediction. See sample API call:
- Feature could be combined with current automated recommendation approach to allow users to modify attributes of the playlist after it has been auto generated on the production application. Example:
Supervised Learning
- Current training / prediction based on unsupervised model because we literally have no labeled data.
- Introduce supervised learning into the DS workflow (possibly using like song activity as the target)
- Would require storing user 'like' activity which would need to be passed from FE to DS
- Can utilize partial fit approach to retraining supervised model once created
Containerization
- Pursue more containerization on AWS ECS for optimizing distributed computing (faster run time)