Former Head of AI at Rappi discusses Model Monitoring in Production
Machine learning models are complex entities. We trust these models with important decision-making and might be trusting them with our lives as autonomous vehicles are starting to come into practice. 
 
However, more often than not, these models are deployed and then forgotten by data science teams after they have moved on to the next project and the company only finds out about a critical error after it has caused significant damage.
 
That is why monitoring ML models when they are fully deployed and running in production is crucial. 
 
In a brief yet informative interview with Wow AI, Mr. Carl Handlin, the former head of AI at Rappi and currently the CTO of Trully, will discuss the importance of monitoring ML models in the manufacturing industry.
 
Watch the whole interview here.
 

About the speaker

 
 
Carl Handlin is a data scientist and machine learning engineer with over 8 years of experience in data extraction, machine learning, modeling, and feature selection and a background in physics and computer science. 
 
Previously the Head of AI for the Latin American region at Rappi, a consumer tech company that specializes in providing online delivery services, he is now the co-founder and CTO of Trully, a collective intelligence identity enrichment platform.
 
Joining the Worldwide AI Webinar this September 29-30, Carl Handlin will be talking about  modern monitoring at scale including some of the current problems that exist when using machine learning or applying machine learning in a real industry.
 

On the vitality of monitoring ML models

 
For Carl, monitoring ML models means closing the cycle. Monitoring means giving feedback to the system, getting feedback from the system, and creating continuous learning, not only to explore the change in the system but also to prevent and anticipate common problems. 
 
He stated that AI/ML changes the input of variables quickly, especially in dynamic settings. Concept drifts, which refer to the unknown and hidden relationship between inputs and output variables, might change the expected behavior of the model. 
 
 
“So being able to monitor and adapt to these types of ever-changing environments is one of the most important things, at least that we know that modern monitoring can help close that gap.” - Carl Handlin, CTO of Trully & Former Head of AI of Rappi
 
Mr. Handlin believes that having a good model monitoring system allows us to find the right time to retrain, adapt, and receive constant feedback throughout the process, which is absolutely essential for not only the developers but also the end-users of the models.
 
 
Carl Handlin will be sharing in more detail about monitoring ML models at the Worldwide AI Webinar. 
 
Grab your free spot: https://wow-ai.com/event
 
Watch the whole interview between Carl Handlin and Wow AI here.