### Purpose
- Create a model for Powerplay based on what parameters they're looking for.
### Principles
- Quality Results - Powerplay has a clear goal and idea on what they want to do. For them it's about getting good results and giving them good accuracy.
### Outcome Visioning
Powerplay is our most important and biggest customer. It's also one where we can really establish the use case. With Powerplay if we can give them a good experience and get good results we can build a strong product case. The hope is that we're able to build them three to four models for retention and acquisition that are valuable to them.
### Brainstorming
- What do I already know about this project?
- I know the code that needs to run and how to clean and structure Mixpanel data.
- I have a few ideas and techniques to run some models. Need to figure out if it works well for them.
- What donβt I know that I need to find out?
- The scale of the users for Powerplay may cause some issues for us. We need to optimise to a certain level. This might involve caching or creating some new backend infra.
- I don't know what exact models they're looking for. More information can be got only after that.
- Who can I talk to who might provide insights?
- What can I read or listen to for relevant ideas?
- Not sure to be honest. Maybe can look into Azure ML cloud a little and understand how people run and train machine learning models on the internet.
### Actions
- [x] Get metadata for mixpanel events for Powerplay π
2023-08-16 β
2023-08-18
- [x] Brainstorm with Bhato on how to create a model π
2023-08-23 β
2023-08-23
- [x] Generate training data set Powerplay π
2023-08-23 β
2023-08-24
- [x] Create a model and iterate on it to arrive at a good model π
2023-08-24 β
2023-08-25
- [x] Create testing data set and test the model π
2023-08-24 β
2023-08-25
- [x] Compute local feature importance π
2023-08-25 β
2023-08-25
- [x] Create backend for B2B SaaS Companies π
2023-08-27 β
2023-08-29
- [x] Create new UI for B2B SaaS Companies π
2023-08-27 β
2023-08-29
- [x] Publish the model results π
2023-08-28 β
2023-08-29