# Meeting Minutes

###### tags: `Meeting`

:::info
- **Location:** [Zoom](https://www.google.com/url?q=https://zoom.us/j/97485649408?pwd%3DeVpyOCtuUVl1YjBZWWY2dmQ5dUJBUT09&sa=D&source=calendar&usd=2&usg=AOvVaw0c_N7XHHseElkOomdrF9-x)
- **Date:** Thursday, August 20⋅11:00 – 11:15am (PST)
- **Agenda**
1. Q&A Chat `30min`
> [name=Hunter]
- **Participants:**
- Morgan
- Fadi
- Camilo
- **Reference:** - [Slack Thread](https://kelahealth.slack.com/archives/C1T6R49GE/p1597860684005000)
:::

### :closed_book: Notes

- Use the AMZN SageMaker examples
- model training
- model serving
- Implement discrete functions for
- training
- model
- containerize your python
- use sagemaker API to deploy algorithm you already built
- need a heads up if using any of the massive instances
- send metrics to cloudwatch
- sagemaker-specific metrics to watch ...

### tooling

- [DVC](https://dvc.org/) instead of github
- [mlflow](https://mlflow.org/)

### naming convention

project folder with in and outboxes is a simple start but gets messy over time.

### :books: Learning Material

[PyData](https://pydata.org/) for the conference talks
- [Data versioning in machine learning projects](https://www.youtube.com/watch?v=BneW7jgB298&t)

### Gaps

- Didn't get the @fadi's list in the meeting agenda in time
:::info
- The current ML workflow (how version control and development outside of SageMaker are integrated, if at all)
- Tips and tricks for those just starting out
:::
- Don't know what specific metrics to track on the models SageMaker generates