Filter (clear filters)






Overview for machine learning

Apache Beam meetup 7 at Datatonic: Beam at Lyft + datalake using Beam + schemas

See how Lyft and Datatonic are using Apache Flink, Apache beam and python in stream processing, machine learning and analytics.


Petastorm: A Light-Weight Approach to Building ML Pipelines @Uber

Data produced and managed by Big Data systems like Apache Spark and Hive cannot be directly consumed by Deep Learning systems like Tensorflow and PyTorch. Petastorm bridges this gap by enabling direct consumption of data in Apache Parqet format into Tensorflow and PyTorch. In this talk, Yevgeni Litvin a senior software engineer with Perception team at Uber Advanced Technology Group (ATG) describes how Petastorm facilitates tighter integration between Big Data and Deep Learning worlds; simplifies data management and data pipelines; and speeds up model experimentation.


Applying Deep Learning To Airbnb Search

This discussion is about the use of machine learning at Airbnb. It's a success story about how machine learning helps Airbnb's search ranking to find guests the best possible options while rewarding the most deserving hosts. Ranking at Airbnb is a quest to understand the needs of the guests and the quality of the hosts to strike the best match possible. The talk discusses the work done in applying neural networks in an attempt to break out of that plateau and also focuses on the elements found useful in applying neural networks to a real life product. 


Scaling a Fantasy Sports Platform with Amazon ElastiCache & Amazon Aurora

Dream11, India’s leading sports-tech startup has a growing base of 40 million+ users playing multiple sports, including fantasy cricket, football, and basketball, and it currently serves one million concurrent users, who produce three million requests per minute under a 50-millisecond response time. In this discussion, Dream11 CTO Amit Sharma explains how the company uses Amazon Aurora and Amazon ElastiCache to handle flash traffic, which can triple within a 30-second response window. Sharma also talks about scaling transactions without locking, and he shares the steps for handling flash traffic—thereby serving five million daily active users.