In its relatively short lifetime, Airbnb
has had over 100 million guests, with over 40 million of those occurring in the past year. Naturally, this exponential growth is a challenge to deal with from a customer service perspective, leading the company to deploy machine learning techniques to improve it.
is a Machine Learning Scientist at Airbnb, where he leads efforts on building a scalable machine learning infrastructure that enables data scientists and engineers to explore, train, and deploy models with minimal effort. He has concentrated on leveraging the vast amounts of text data on the site to enable the next generation of data products within the company.
At the Machine Intelligence Summit in New York
, Avneesh will share expertise on extracting insights with machine learning, concentrating on how customer issues can be addressed in a shorter amount of time, while maintaining a high level of user satisfaction. Customer issues in aggregate can be revealing for potential product improvements and changes, and Avneesh will discuss how they at Airbnb extract and categorize these potential improvements from vast amounts of service tickets.
I asked him a few questions ahead of the summit to learn more.
What are you currently working on at Airbnb?
I concentrate on deploying natural language processing models in a variety of domains across the company. I also work on building a common, scalable machine learning infrastructure that empowers data scientists and ML-oriented engineers to explore, train, and deploy models with minimal efforts.
What do you feel are the leading factors enabling recent advancements in natural language processing?
The use of word embeddings to succinctly capture lexical semantics and improve a wide variety of downstream tasks in text classification, shallow and deep parsing, and machine translation (amongst others). NLP has traditionally suffered from a data sparseness problem (primarily because there are a potentially infinite number of ways to express a single semantic concept) and word embeddings have helped with this problem considerably.
What present or potential future applications of NLP excite you most?
Real-time speech to speech translation. I think we’re still nowhere near a successful technology, but the potential to translate speech from one language to another is what excited me about the field in the first place.
Which industries will be most disrupted by Machine Intelligence?
Practically every industry! From a slightly biased perspective though, I think 21st century travel will be significantly more personalized and bespoke (hopefully with Airbnb leading the way!), and machine intelligence will play a significant component in this endeavor.
What developments can we expect to see in Machine Intelligence in the next 5 years?
From a research perspective, deeper theoretical understanding of neural networks and non-convex optimization in general, and also additional work in semi-supervised neural network training. From an applications perspective, we will get used to the concept of a personal digital assistant constantly optimizing and tweaking our lives based on patterns from previous days.
Avneesh Saluja will be speaking at the Machine Intelligence Summit in New York, on 2-3 November. To register, visit the event website here.Other speakers at the summit include Kamelia Aryafar, Senior Data Scientist, Etsy; Naveen Rao, CEO & Co-Founder, Nervana Systems; Carl Vondrick, PhD Student, MIT; Tara Sainath, Senior Research Scientist, Google; Kathryn Hume, President, Fast Forward Labs; Antoine Bordes, Research Scientist, Facebook AI Research; and Siddartha Dalal, Chief Data Scientist, AIG.See the full events list here for events focused on AI, Deep Learning and Machine Intelligence taking place in London, Amsterdam, Boston, San Francisco, New York, Hong Kong and Singapore.
Deep Learning Algorithms
Machine Intelligence Summit