Leveling up your ML Infrastructure

Details
For many companies the manual and data-scientist-driven processes might be sufficient when models are rarely changed or trained. But whenever it is not enough the automation process of using new data to retrain models in production comes into play and companies need to revisit their ML Infrastructure. In this session we will cover MLOps fundamentals, present some common challenges during the design process and discuss workarounds when other scenarios are not appropriate.

Agenda:

1. Maturity levels of MLOps platforms. Where are you?
2. Feature stores. Do we really need them?
3. Flow management tools. MLFlow, Azure and more
4. Deploying the "Deployment Pipeline" instead of the model instance
5. Target detection mechanisms for performance monitoring and common workarounds

About the host
Lilit is a Data Scientist in ServiceTitan. ServiceTitan is the world’s leading and fastest-growing software technology platform for the trades, a trillion-dollar global industry. We help small business entrepreneurs run and grow their businesses streamline operations, improve customer service, and increase sales. It currently has more than 4,000 customers and more than 1000 employees across Glendale, Atlanta, and Yerevan.

Join Zoom Meeting
https://us02web.zoom.us/j/89496269970?pwd=NjdXMjRvajA2VGRLcXdsRGFReXdadz09

Meeting ID: 894 9626 9970
Passcode: 999508

To Join ADSC Slack: https://join.slack.com/t/armenian-ds/shared_invite/zt-lztlm1d1-EMJTcJmuaHa4637T4WnbXQ