AI tech talk — Big data and Personalization using machine learning
Samik Raychaudhuri, director of Data Sciences Group at 7.ai, gave a talk on big data and personalization using machine learning at Arvind Internet office.
The session covered various aspects of personalization right from basics of personalization, evolution of it, types of personalization and where all it is used in industry. How companies in various business domains are using user based personalization to improve end user experience. How to implement personalization in-general, typical algorithms, tech stack and measure accuracy of the results.
Then he went on to cover how his team at .7 is using big data and deep learning to transform their business, solving complex problems and doing real time personalization given huge volume of data they have.
A very engaging session, followed by QnA from AI team.
Some of the takeaways from his sessions:
- 70% of total time will go in cleaning, converting unstructured to structured form, deal with missing data and overall data preparation challenges.
- Good way to start learning about machine learning is to take a mid-size problem that can be implemented in the production. Start around basic ML models and collaborative filtering.
- DL4J is one good framework for implementing Deep Learning in java.
- In consumer facing interactive AI applications (e.g. Chat bots), most important thing is to begin by detecting intent of user
- When working on model, identify features or variables that indicate signals and keep iterating the model.
- What are typical models used for personalization.
Some of the pics of the event: