Agenda 2020

February 27, 2020

The conference agenda is almost ready. There is morning and afternoon plenary session, 5 simultaneous tracks and roundtables discussions session. Plus of course evening networking party!

11.15 – 15.30 Simultaneous sessions

16.00 – 17.25 Roundtables sessions

Parallel roundtables discussions are the part of the conference that engage all participants. It has few purposes. First of all, participants have the opportunity to exchange their opinions and experiences about specific issue that is important to that group. Secondly, participants can meet and talk with the leader/host of the roundtable discussion – they are selected professionals with a vast knowledge and experience.

There will be 2 rounds of discussion, hence every conference participants can take part in 2 discussions

16.05 – 16.45    1^st round

16.50 – 17.30    2^nd round

 Snorkel Beambell – Real-time Weak Supervision on Apache Beam 

1. Deep Learning models have led to a massive growth of real-world machine learning that allows Machine Learning Practitioners to get the state-of-the-art score on benchmarks without any hand-engineered features. 2. The challenge with continuous retraining is that one needs to maintain prior state (e.g., the learning functions in case of Weak Supervision or a pretrained model like BERT or Word2Vec for Transfer Learning) that is shared across multiple streams. 3. Apache Beam’s Stateful Stream processing capabilities are a perfect match to include support for scalable Weak Supervision.

Real-life machine learning at scale using Kubernetes and Kubeflow

How to build a machine learning pipeline to process 1500 TB data daily in a fast and cost-effective way on Google Cloud Platform using Kubeflow? How to serve TensorFlow model with almost 1M requests per second and latency < 10ms on Kubernetes? Is Kubernetes and Kubeflow ready to serve data scientists?

Managing workflows at scale

How to build and maintain thousands of pipelines in the organisation? What are the biggest pain points in orchestrating hundreds of ETLs? What open source and managed solutions are available?

Practical application of AI

Industry 4.0 and AI – are we ready for the 4th industrial transformation? Who should be the beneficiary of the Industry 4.0? Key barriers to the implementation of AI projects in organizations. Real cases of AI in Industry.

The need for explainable AI

With the spread of AI-based solutions, more and more organizations would like to understand the reasons for system decisions. It’s especially interested in regulated industries. The session will cover so-called white-box methods, as well as modern approaches to AI explainability which allow understanding more complex models.

A.I. and his vulnerabilities

Nowdays AI and ML are everywhere as long as look like your project is not cool without them, they are able to to outperform human performaces in some tasks
but can you always trust the math behind? Do we really need A.I. & ML always?
Open discussion on vulnerabilities and easy hacks intrinsics in math models and the impact in real life scenarios and our security.

Deploying ML models in real-time using stream processors

Today, streaming engines enable data pipelines to continuously deliver AI-based outcomes towards what is frequently called continuous intelligence. Which are the most important architectural choices when latency comes as the top requirement? How can we scale thousands of models in production? In which ways realtime environments help ML model evolution with zero downtime?

Data discovery – building trust around your data

Worldwide growth of data has changed business landscape forever. Multiple organizations undergo transformation triggered by the data revolution. While one can understand the benefits of collecting bigger data volumes, it has revealed additional challenges when trying to effectively use it. Ability to explore the data and increasing compliance demands force us to think about solutions to leverage power of metadata. Data description evolve from being a simple schema definition to catching application context, behavior and how it is changing over time.
Let’s discuss data discovery in context of use cases, technologies and possible challenges.

What to do with my HDP/CDH cluster with new Cloudera licensing model

After the merger with Hortonworks, Cloudera becomes a single vendor that builds a distribution that consists of major components from so called Hadoop Ecosystem (e.g. Hadoop, Spark, Hive, Ranger). While these components itself are open-source, access to binaries that are critical to install/upgrade the clusters will be limited to only customers who purchase a paid subscription. This means that thousands of the companies that currently use Hadoop for free, will need to decide what to do next. Should I pay for a subscription or compile own binaries to build own distribution? Should I stop using on-premise Hadoop and go to the public cloud instead? During this panel we explore this topic and try to answer these questions based on our vendor-neutral experience when working with our customers who have large production installations of HDP/CDH clusters.

Being efficient data engineer. Tools, ecosystem, skills and ways of learning

What does it mean to be a productive (data) engineer? Is it about the tools we use? Is it the mindset we have? Is it the environment we are surrounded by? Let’s share and discuss war stories, learning resources, methodologies and libraries that help us escape the gumption traps in the daily life of an engineer. Discussion will be divided into 4 areas: debugging, implementation, communication and learning.

Databases in Kubernetes: from bare metal to cloud native

Initial cloud-native conquests started with stateless services but gradually turned towards data management systems. DBMS have relied on centralized, bare metal servers for decades. Cloud-native architectures are a big new technology shift for such systems. Many commercial and open source databases already provide cloud-native adoptions of their products. Among the most interesting cloud-native converts are analytic databases, which imply additional requirements to storage and clustering techniques in order to run fast analytic queries over billions, or even trillions, of rows. Examples include the MySQL clustering project Vitess, which has recently reached CNCF graduation level, and ClickHouse, an extremely fast and scalable analytical database that is being converted to cloud-native operation by Altinity.

Join me to discuss various aspects of running databases in Kubernetes. This is a new technology that has a lot of caveats, such as storage. At the same time databases in Kubernetes promise substantial benefits to the users of such applications as well as companies that operate them. We will explore these issues as well as the path to maturity.

Handling Data Science experiments efficiently – methodologies, tools, collaboration

A typical data science project requires usually a lot of playing around with the data. What’s the best method to keep track of all the experiments within in a team? Using a spreadsheet, custom tool or recently popular MLFlow, Neptune or maybe other tools? Let’s talk about what works best depending on the type of a project, a team’s size and other aspects.

Managing a Big Data project – how to make it all work well?

How to ensure successful adoption of Big Data and Analytics systems? It is a challenge for most organizations. Let’s discuss how to promote user-centric approach, leverage experience design and manage user expectation on Big Data projects. I would be happy to hear you opinion and answer your questions, based on my practical experience applying Design Thinking and architecture design methodologies. I believe this conversation will be interesting for Architects, Tech Leaders, Product Managers and C-level folks.

Analytics and Customer Experience Management on top of Big Data

How to ensure successful adoption of Big Data and Analytics systems? It is a challenge for most organizations. Let’s discuss how to promote user-centric approach, leverage experience design and manage user expectation on Big Data projects. I would be happy to hear you opinion and answer your questions, based on my practical experience applying Design Thinking and architecture design methodologies. I believe this conversation will be interesting for Architects, Tech Leaders, Product Managers and C-level folks.

Challenges of building a modern & future-proof data processing platform

The speed of changes in IT and our companies seems to never stop increasing, especially in the field of BigData. To keep up with it we need to move fast and be smart about it, but how to actually achieve that? How to predict future needs for processing and tools? How to prepare for it? What kind of trade offs we can make? We’ll try to answer that together and share good practices and experience during this session.

Serverless data warehousing

Definition what is serverless warehouse. List of solutions which are considered serverless. Data ingestion. Data storage. Data processing, pricing and cost efficiency. Advantages and disadvantages of both serverless and on premise.

Scale Your Logs, Metrics, and Traces with the Elastic Stack From traditional applications to microservices and Kubernetes

How do you tackle your monitoring and observability problems? There is a high chance that you are using the Elastic or ELK Stack and this session is all about making it scale: From easier collection of data, to scalable multi-tier architectures, and the lifecycle of your data including the deletion.

Full-Text Search with Elasticsearch and Apache Lucene — Finding What Is Relevant “You know, for search” is the tag line here

With ever increasing data volumes, finding what is relevant in a performant way is a common requirement. This discussion is for common problems, recommended architectures, best practices, and new developments around the most widely used search technologies.

SQL on Big Data for batch, ad-hoc & streaming processing

Bring Data as Products to consumers

How do you define data products? What mindset and approach we should have to make Product approach possible in Data? Let’s brainstorm about implementation of the approach, opportunities and values it brings.

Data visualization, how to visualize large, complex and dirty data and what tools to use

Addressing challenges of modern analytics with Snowflake

Stream processing engines – features, performance, comparison

From on-premise to the cloud: an end to end cloud migration journey

Elasticsearch for search of logos, metrics and user-data

Choosing and using the right BI platform for efficient data exploration, reports and analytics

Beyond pre-computed answers – interactive, sub-second OLAP quires with Druid, Kylin and Clickhouse

Big Data on Kubernetes

The Latest and Greatest of Apache Spark

Best tools for alerting and monitoring of the data platforms

Data Auditing – How to get a clear view of your pipeline

Optimizing Spark-based data pipelines

Despite the spread of dedicated AI platforms, ready-to-use ML libraries and tons of data available, running successful large-scale AI/ML projects still faces technical and organizational challenges. According to some studies, 8 out of 10 such projects fail. This panel will explore the necessary technical prerequisites that a company should introduce to build ML-based solutions efficiently. This includes, for example, organizing the data (e.g. data discovery, data lineage, data quality), experimenting with models (e.g. notebooks, libraries, collaboration), one-click-deployment of a model (e.g. AI/ML platforms, infrastructure) and more. While many of these challenges are not that hard when working with small data, everything gets more complex & time-consuming when the data and scale is larger.

At the end of the conference we would like to invite all the attendees for the informal evening meeting at “Dekada” Club , which is located at the Grójecka 19/25, 02-021 Warszawa.