Machine Learning in Production : Fundamentals and Updates
< Work in Progress >
When both technology and ecosystem are rapidly evolving, one of the prerequisites to excel is to focus on building things that either lasts longer or truely differentiates itself amongst currently available alternatives. If you are a Machine Learning practitioner, it’s not hard to end up in a situation where several research papers and prototypes of a new algorithms are out on standard datasets before even your older prototype makes it to production on real dataset. However, newer algorithms can become least of your concerns if fundamental plumbing, raw materials (data) and measurement metrics of system are set up correctly. This talk will focus primarily on setting up few key components of machine learning pipeline correctly such that you can efficiently and cost effectively move your prototypes to production.
< Work in progress >
Introduction [2 mins]
Clarify all Ws before starting (Why, Who, What) [5 mins]
See what is available already out there and ready to use. (Some latest updates in the ecosystem to set the stage for rest of the talk)[5 mins]
Garbage in garbage out (Let there be enough data! Data is expensive. How to optimise on training data generation?) [7 mins]
Modelling (Simple heuristics – > simple model –> complex model, Launch and iterate, What’s the latest stuff?) [6 mins]
Integration and Deployment (In memory or micro service or lamda or on device?) [6 mins]
Conclusion [5 mins]
Machine Learning enthusiasts who wish to be good engineers :P
Krupal started his career as a Research Trainee at Hewlett-Packard Laboratories 6 years ago and currently leads all Machine Learning initiatives at Haptik. He is also recognised amongst less than 50 Google Developer Experts globally in Machine Learning for his open source contribution, technical blogs, public speaking and mentoring in the community. He specialises in rapid prototyping of machine learning algorithms and has efficiently deployed multiple models to production addressing different use-cases. He likes to mentor engineers and researchers to help them align their efforts in result-oriented direction. A part from technology, he also likes to read and learn about product development and business strategy. One of his dreams is to solve a real world problem which positively impacts atleast one of the fundamental needs of human race.