You have all these documents - docs, emails, PDFs, forms, images… They could give you valuable insight into your business and customers to help you make better decisions - but instead most documents are just sitting there, untapped, because it’s difficult to read, compare and understand the relationship between them.

#ArtificialIntelligence

#DocumentAI

#ComputerVision

#NLP

Diverse algorithms have been developed for efficient earthquake signal processing and characterization. These algorithms are becoming increasingly important as seismologists strive to extract as much insight as possible from exponentially increasing volumes of continuous seismic data. Deep neural networks have been shown to be promising tools for this. We have developed a number of deep learning tools for more efficient processing and characterizing of earthquake signals. In my presentation, I demonstrate the performance of some of these tools applied to seismic data. AI-based techniques have the potential to improve our monitoring ability and as a result, understanding earthquake processes and hazards.

#earthquakemonitoring

#AI4earth

#earthquaketransformer

#scientificmachinelearning

#dataengineering

#management

#datateams

#datascienceoperations

During the session we will focus on the AI Powered Banking with Backbot,  AWS Lex Chatbot for everyday banking and the real world limitations of using AWS lex in banking environments. I will present a prototype I created for Backbase to show how we can leverage AWS Lex for our customers to perform basic tasks like getting your account balance, make transfers between your own accounts dan to your beneficiaries.

The presentation will discuss the real life implementation limitations and constraints when working with AWS Lex

#aws

#lex

#chatbot

#banking

#nlp

Text summarization is a highly useful tool for extracting key information from text, which helps speed up learning, communication and business processes dramatically. With the use of modern attention-based architectures, automatic production of human-like summaries has become easier to achieve than ever. For some of the inputs, however, the quadratic time complexity of transformer encoders results in a specific token limit, acting as a constraint for processing large text sequences. In her talk, Nataliia will show how to produce high quality summaries for large text inputs, guiding you through some of the smart methods to deal with the token limit. She will compare the performance of different state-of-the-art methods on large text sequences on the example of consumer complaint data and show which algorithms ensure producing the best summaries.

#AI

#ML

#IoT

#WebAssembly

#Deployment

AI technology is creeping into every industry. In the networking industry, the proliferation of devices, data, and people has made IT infrastructure more complex than ever to manage, with many looking to AI for help. Together AI and ML play an increasingly critical role in taming complexity for growing IT networks.

#AIOps

#MistAI

#Marvis

#UserExperience

#ClientVisibility

In each journey of implementing advanced analytics in an organization, there comes a moment when POCs turn into projects, and projects as an integral part of business processes. Managing the day-to-day maintenance of these processes is not easy because of the ever-changing nature of Machine Learning. How to maintain control over the changing versions of models, data sources and processes? In the session, I will show you how to reflect the full Data Science Lifecycle in Azure and set up a mature MLOps process for it.

#MLOps

#onboardprocessing

#deeplearning

#hyperspectralimageanalysis

#Earthobservation

#satelliteimageanalysis

#imageduplicates

#opensource

#computervision

#python

#AI

Rapid development of advanced modelling techniques gives an opportunity to develop tools that are more and more accurate. However as usually, everything comes with a price and in this case, the price to pay is to loose interpretability of a model while gaining on its accuracy and precision. For managers to control and effectively manage credit risk and for regulators to be convinced with model quality the price to pay is too high. So, it prevents them from using advanced models due to the lack of their interpretability. In this paper, we show how to take credit scoring analytics in to the next level, namely we present comparison of various predictive models (logistic regression, logistic regression with weight of evidence transformations and modern artificial intelligence algorithms) and show that advanced tree based models give best results in prediction of client default. What is even more important and valuable we also show how to boost advanced models using techniques which allow to interpret them and made them more accessible for credit risk practitioners, resolving the crucial obstacle in widespread deployment of more complex, “black box” models like random forests, gradient boosted or extreme gradient boosted trees. All this will be shown on the large dataset obtained from the Polish Credit Bureau to which all the banks and most of the lending companies in the country do report the credit files. This huge extent of data ensures high quality of the model inputs and objectivity of the conclusions.

#XAI

#CreditScoring

#AI

#ML

#Twolayerappraoch

#CombineArtificialandHumanIntelligence

Germany is at the forefront of Machine Learning (ML) research in Europe whether it is published papers or issued patents in the past decade. As a consequence, theoretically you would expect a lot of ML in our products across all industries and a certain maturity of the field in Germany. The reality is however different. ML is a “Neuland” for many companies here and we are not good at taking state-of-the-art research results into production. In this talk, I will present why this is the case and then show some recipes on how to solve this problem.

Connecting data to context means to have the right data, use the right algorithms and deliver the right insights in a way they play an integral role in the insurers and customers daily life. The presentation is showing how to support with data products with a focus on lead management in the insurance industry. To go the full journey of a data product the use case, the data and model evaluation, and most important the visualization and implementation will be shown using a real example. In recent years, many companies have begun to invest in AI. However, most AI projects hardly make it beyond a proof of concept and thus do not contribute to increasing competitiveness. This talk shows how to identify AI potentials in companies. We will also discuss which managerial and technical challenges can arise when implementing AI applications. You will learn a number of best practices when implementing AI and how these lead to productive applications that ultimately add financial value.

#predictiveanalytics

#conversionmodeling

#leadpriorisation

#dataproductimplementation

#visualization

The problem: How food waste negatively affects our climate
The solution: “Tech for Good” - Using an AI driven distribution platform to match time-critical perishable food oversupply with the demand
The approach: Meeting and understanding the needs of supply and demand partners
The impact: Using existing food resources using a circular economy approach, reducing unnecessary CO2 emissions (environmental impact), supporting people in need and NGOs (social impact)

#techforgood

#reducefoodwaste

#climateaction

#circulareconomy

#SDGs

#data

#engineering

#dataengineering

#modenstack

#scaleteams

A classified website is an online marketplace where people can sell and buy new and used items from a wide selection of categories. The vast majority of user actions on these platforms is done between buyers and sellers without any real payment transaction. Therefore, as eBay Classified Group (eCG), we had to come up with an operational definition of conversion: if a buyer sends a message to a seller or if a seller posts a new listing on the platform, we refer to them both as "converted". The eCG is an umbrella company managing 14 different classified platforms from all across the world. Conversion numbers for the group company are tracked by a central team named 'Global Growth', which is tasked with increasing the number of active (recently converted) users on the platforms. They turned to the Data Science Team with a question: "Who will have a conversion soon?" The initial purpose of this model was to let them know in advance who is likely to convert soon, so that the marketing team would be able to take actions on users beforehand. The idea is straightforward: focus only on users who are unlikely to convert soon and try to find different ways to convince them to get back on our platform again (also: avoid allocating resources on users who will convert soon anyway, without requiring any extra targeting). This personalised marketing strategy was the main objective of this project. This is an article about an ML model developed to accomplish this objective by predicting user likelihood of conversion in the near future based on the users' past actions on a platform.

#MachineLearning

#BigData

#PredictiveModeling

#ConversionRateOptimization

#PersonalisedMarketing

There have been several evolutions ongoing around data handling. First, we saw the datawarehouse. This later led to the second evolution - the datalake. Now, we see the next trend called "Data Mesh". What is the data mesh and how does it line up with the previous ones? What is different and why is it not a technology? Get the answers in this session.

The WiTraPres project (conducted by the data science lab at the FH Südwestfalen, Germany) aims at implementing a predictive maintenance module to a fully automated warm reshaping plant (which produces parts for vehicle interiors) using machine learning methods. More precisely, insights gained from data are to be combined with expert knowledge in order to achieve predictions that are as accurate as possible. In the data science component, new predictive maintenance methods are developed and applied using process data and are then compared with conventional methods. One approach aims at implementing Bayesian-LSTM-Autoencoders for anomaly detection. Here, Bayesian methods are coupled with time-resolving deep learning methods to achieve the highest possible predictive power of the model with high precision. The predictions made by this model can be effectively combined with expert knowledge. Our approach shows promising results for this case specifically and is able to compete with state of the art predictive maintenance methods.

#machinelearing

#deeplearning

#predictivemaintenance

#automotivemanufacturing

#kubeflow

#pipelines

#bert

#nlp

#sagemaker

Deep learning approaches have proven powerful in modelling the volatility of financial stocks and other assets, as they are able to capture non-linearities in sequential data. I will present an analysis that helps asset managers to select, forecast and analyse different optimal asset portfolios over multiple backtest iterations. - The analysis includes the preselection of stocks using an autoencoder model, a new way to clean the sample covariance matrix that describes the risk of a portfolio, a ten day forecast using recurrent neural networks and the final portfolio optimization.

#DeepLearning

#QuantativeFinance

#Autoencoders

#PortfolioOptimization

We investing a lot of effort on building data teams, scalable architecture, data engineering, BI, analytics and Machine learning. But are we doing the right things? How much are we actually contributing to the organisation? Are we focusing on the right things? In this session I will share my experience on how to measure success in data teams and the benefit of doing so.

During the Covid 19 pandemic, the volume of requests via digital channels has increased massively. One solution to ensure excellent service without overburdening staff:in: AI-based chatbots. Intelligent bots can handle standard inquiries quickly and automatically. However, this does not always work properly: almost 92 percent of interactions with bots in Germany contain swear words and insults.

Jens Leucke is therefore happy to explain:
• for which areas of application AI-based chatbots are currently being considered and for which they are not
• which technological aspects and functions are particularly relevant for chatbots
• which strengths and weaknesses AI currently still has in this area - and how the technologies can be further improved
• which steps need to be taken when implementing AI-based solutions in customer service in order to significantly improve the user experience

Everybody talks about artificial intelligence and machine learning these days, but how many of the solutions mentioned genuinely involve a **machine** that is **learning**? There are multiple results being published about e.g. DeepMind beating Starcraft - and most Atari games have been mastered by RL algos a while ago - but about actual business applications? In this talk, I will give an overview of how you can apply associative reinforcement learning to a real problem: online content personalisation (and stay within budget, because open source).

Creating the future of remote maintenance is strongly connected with aerial applications of drones. In the industries of energy, construction and facility management, drones are a fully established tool for various tasks. FlyNex enables companies to automatically generate and collect data. By managing the data within the FlyNex Platform companies can analyze them with a chosen (AI-based) software, which can be fully integrated into the platform.

#aerialremotemaintenance

#drones

#AI

#FlyNex