Data Science Lightning Talks 04

The next Lightning talks will take place on June 7, 2022. As usual, the meetup will consists of several very short talks covering various data science topics. The idea of lightning talks is to give more speakers an opportunity to present and to get a better overview what kind of data science is getting practiced.

Agenda:
17:30 - 17:35 - Intro
17:35 - 18:35 - Lightning talks
18:35 - 19:30 - Discussion & refreshments

Talks:

ML Ops: Machine Learning Operations by Bostjan Kaluza (Chief Data Scientist @ Evolven)
To complete a successful machine learning project, understanding machine learning theory and concepts is essential, but you need engineering capabilities as well. This talk will highlight topics how to plan, build and maintain a machine learning project and share a plan on organizing a series of workshops on this topic next study year @ FRI.

How Genialis makes big data from small datasets to advance cancer treatment by Miha Štajdohar (Co-founder and CTO @ Genialis)
Genialis’ mission is to make sure today’s most promising therapeutics reach patients of the highest unmet need. We use the knowledge learned from disease biology to predict patient response to treatment. The challenge is clinical datasets are typically small. The solution is to model the biology and validate the classifier across multiple datasets. We will present the Data Flywheel for AI/ML, a suite of technologies to integrate and manage disparate “small” datasets that together can be used to develop and validate clinical biomarkers (predictors of patient response) in a regulated environment.

Clustering with imperfect distance metrics by Rok Piltaver (Senior Data Scientist @ Ekipa2, a subsidiary of Outfit7)
To cluster objects, we need object data, a clustering algorithm, and a distance function. But how to do clustering if there's no good distance function due to missing domain knowledge or high complexity? We'll discuss a potential solution that perform clustering in several steps with partial distance functions. I'll demonstrate how we used this approach to cluster army formations in Mythic Legends, a new auto-chess inspired strategy role-playing game. We learned from snapshots of real user formations to generate formation templates for high quality non-player characters.

How data enables the Metaverse by Ivan Varko (CEO @ InfoMediji)
Technology is heading towards the next level of abstraction and virtualization making the Virtual Reality Metaverse come true. At InfoMediji we are deploying a VR video streaming pipeline enabling creators to connect with millions of early VR users all over the world. My talk will be a brief discussion how VR works, the problems and challenges we face, and an open invitation to anyone who wants to contribute to the creation of the Metaverse.

From pies to impact: design in data visualization by Tina Lekše (Data Engineer @ADD, Student @ Business Informatics (SEB LU))
As we all know, data driven business decisions are what business intelligence strives for, but nonetheless, having a dashboard showing you irrelevant information can lead from useless to wrong decision making. In this short talk, I'll try to go quickly through some data visualization design and color principles, which can make or break the story you are trying to tell with the data. We'll discuss design, color, and chart types, and what to keep in mind the next time you're building your BI dashboard.

Lakehouse in 7 minutes by Jure Jeraj (Head of Data Engineering Team, Big Data Specialist @ Result)
Lakehouse is quite new construct in data ecosystem. It supposed to be Data Lake with DataWarehouse (transactional) abilities. Initially was introduced by Databricks but main architecture is really general. In just 7 minutes, I will present you main reasons and logical elements of Lakehouse architecture and which benefits bring to Data Scientist.

Correlation, Causality and the do-calculus by Paul Larsen (Head of Data Analytics Practices @ Allianz Insurance)
We all learn in introductory statistics that correlation and causation are not the same thing, and then typically ignore this lesson when doing data science and machine learning. In this talk, we give motivation for why causality should be taken seriously, an introduction to Judea Pearl's do calculus, and an insurance-inspired example of how causal reasoning leads to different business decisions.