Jai Ranganathan (KeepTruckin)

Jai is currently SVP Product at KeepTruckin, and was formerly VP of various AI and Data matters at Uber.

End-To-End Use Case of Uber's COTA system

Uber's Customer Obsession Ticket Assistant (COTA)

• A tool that uses machine learning and natural language processing techniques to help agents deliver better customer support.

• Enables quick and efficient issue resolution for more than 90 percent of Uber's inbound support tickets.

Challenge

As Uber grows, so does the volume of support tickets

• Millions of tickets from riders, drivers, and eaters per week

• Global-scale of serving 600+ cities

• Thousands of different types of issues users may encounter

• Multilingual support

https://eng.uber.com/cota/

Customer Support Platform

• Steps in the workflow

  • User → Select Flow Node → Write Message → Contact Ticket → Customer Support Representative → Select Contact Type → Lookup Info and Policies → Select Action → Write Response Using a Reply Template → Response → User

• Problems to solve

  • Issue prediction

  • Issue categorization

  • Ticket routing

  • Ticket volume

  • Policy optimization

  • Auto-response

Exploration

• Identify the right problems to solve

  • Use analytics to understand the value before all else

  • Know what metrics to optimize for

• Understand whether Machine Learning is a good fit

• Build with an eye on the probabilistic nature of Machine Learning solutions

Development

• Many possible solutions including basic Machine Learning techniques

• Understand the cost-benefit of compute time vs accuracy

• Deep learning is a fast-evolving space - keep up with the literature to understand the latest advances

• Validate your results with visualization

Deployment

• Architecture complexity with feature engineering and training have special needs

• Deep learning is still slow! Distributed deep learning can help a lot and is getting better

• Good experiment design required to validate the models

Monitoring

• Dynamic business problems require retraining strategies with well thought out safe deployment

• Continuous improvement of labeling will make your models better

• Look for edges where your models fail to find room for model improvements