JDA Labs and Google Cloud: Data Streaming and Machine Learning Meet-up Recap

JDA Labs and Google Cloud recently co-hosted a meet-up at the JDA Labs office in Montreal. The meet-up focused on data streaming and machine learning with the Google Cloud Platform, and was a very hot topic. This was the most popular meet-up hosted to date by JDA Labs, with a full house and an attendee waiting list.




Check out the recap of the presentations from the meet-up below:

  • The first topic was, “Machine Learning through Google Cloud Platform’s Dataflow”

Maxime Legault-Venne from JDA Labs talked about the goal to predict how well an item will sell based on past performances of similar items. To do so, the development team trains on existing data to understand what attributes have which impact on the performance of items. They use trained models to predict how unknown items would perform based on their attributes. This training process is great but running it takes hours and hours. To reduce execution time and improve the use of resources, JDA leverages Google Cloud Platform’s Dataflow to parallelize this processing in a reasonable amount of time. JDA Labs tested 700 combinations of hyperparameters and training entries and passed from a 12-hour execution time to less than 28 minutes using Apache Beam and GCP Dataflow with their data.

  • The second topic was, “Tweets Have Feelings Too: Building a Dataflow Streaming Pipeline for Sentiment Analysis on Twitter”

Google Cloud’s Arsho Toubi showed attendees how to run a Dataflow pipeline to perform sentiment analysis of Twitter posts using the Google Cloud Natural Language API, then query and visualize those results.

Google Cloud Platform’s Dataflow is a unified programming model and a managed service for developing and executing a wide range of data processing patterns including extract, transform and load (ETL), batch computation and continuous computation. Because Dataflow is a managed service, it can allocate resources on demand to minimize latency while maintaining high utilization efficiency. The Dataflow model combines batch and stream processing so developers don’t have to make tradeoffs between correctness, cost, and processing time.


After the presentation, the attendees stayed an extra hour to participate in discussions with Arsho Toubi and JDA Labs associates.

If you missed the presentation, check out the following resources:

Stay up-to-date with JDA Labs meet-ups! The next meet-up is taking place in January. Don’t miss the chance to reserve your spot here.

About JDA Labs

Ever wonder what the supply chain of the future will look like? What technology advancements will reshape the way we do business? We do, every day.

Looking toward the future is the job of JDA Labs, a global research and development team headquartered in Montreal, Canada. Every day, the creative thought leaders at JDA Labs are focused on delivering game-changing new products, winning patents and developing revolutionary best practices that will shape the future supply chain — and deliver incredible results for our customers. Learn more about JDA Labs here.

No Comments

Be the first to start a conversation

Leave a Comment