confirmed talks so far
On Board Artificial Intelligence : Train, Deploy and Use Deep Learning on an Edge Device, a Raspberry Pi
Machine learning applications are new in the software development landscape, and tend to be hard to build. As Google noted in an article, it is mainly because the application is much broader than the model itself. Surprisingly though, Machine Learning applications follow a double Pareto’s law. On the one hand, 80% of the time spent on building those applications deals with machine learning problems whereas 20% of the remaining time is spent on…
Distributed Deep Learning with Keras and TensorFlow on Apache Spark
DeepLearning4J is an Open Source distributed framework for Deep Learning on the JVM. It allows importing Python (Keras and TensorFlow) models in order to train them in a distributed fashion on Apache Spark. The talk would walk through the reasons for doing distributed Deep Learning of Python models in a JVM based environment and the details to productionalize this process.
Realtime Data Pipelines Using Spark Streaming
At Outbrain we serve billions of personalised recommendations.
Our serving ML models were built on top of batch ELT flows.
But having near realtime inputs is extremely important in our business.
During this session, I will present our journey from batch-based to real-time analytics.
Data Science for Lazy People, Automated Machine Learning
Data science is fun, right? Data cleaning, feature selection, feature preprocessing, feature construction, model selection, parameter optimization, model validation – oh wait – are you sure? What about automating 80% of the work even doing better choices than you? Automated Machine Learning has arrived to be your personal assistant in Data Science.
What It Takes to Build Production Ready AI Solution
We are data company that works with other companies to help them build AI solutions. We are a blend of data scientists and data engineers and that makes us question from different angles how next big AI module will be integrated in your platform. This is prime reason why we can brag that we have more then 10 AI solutions in production developed over the last 3 years.
Data Science & AI: Infrastructure Matters
Artificial intelligence is increasingly being seen as a competitive advantage and every company is running fast to try and be a part of this revolution. However, on its own, AI faces a steep time-to-value curve. For example, do you have the right data? Do you have the right skills? Can everyone participate (typically, AI is in the hands of the privileged few and companies struggle to democratize it for the many)? But there is one often overlooked…
Breaking Out the Lead Scoring Algorithm
The Lead is considered to be any individual who may become a potential client as it has shown an interest in the product or service a company offers.
A Lead Generation refers to the method of collecting leads in order to manage sales channels more efficiently, raise brand awareness and contribute to rising profits. Lead Scoring includes assigning certain weights to each potential…
Visual ML with TensorFlow.js
From Spark to Elasticsearch and Back - Learning Large Scale Models for Content Recommendation
Serving tens of billions of personalized recommendations a day under a latency of 30 milliseconds is a challenge. In this talk I’ll share our algorithmic architecture, including its Spark-based offline layer, and its Elasticsearch-based serving layer, that enable running complex models under difficult scale constrains and shorten the cycle between research and production.
Predicting the Moment of Calving in Dairy Cows Using Time Series Analysis in Apache Spark
Stillbirth, defined as calves that die during unobserved birth is often seen as an indicator of lowered animal welfare in dairy cows. Sensors have been proposed as a tool to support dairy farmers but accurate calving prediction models are often lacking. In this session, a machine learning data pipeline will be described using the spark ML framework. Heavy lifting and feature preparation for sensor data from 1331 cows on 8 herds from 21 days before until the day of calving was performed using sliding windows and time series analysis. In total a set of 100+ features was used in a random forest classification model trained and tested using different cross validation approaches.