Zapraszamy na trzydzieste spotkanie grupy PyData Bydgoszcz! Oprócz tradycyjnej prelekcji zobaczycie także SPOCK (Studenckie Prezentacje o Ciekawych Kwestiach). W ramach tej serii, studenci z koła naukowego SOMSIAD zaprezentują swoje odkrycia technologiczne w 15-minutowych prezentacjach!

Kiedy: czwartek 12.12.2024, 18:00-20:05 (ROZPOCZYNAMY PUNKTUALNIE!). Gdzie: Młyny Rothera (Spichrz Zbożowy, 2. piętro), Mennica 10, Bydgoszcz. Cena: zupełnie za darmo! Do tego konkursy z cennymi nagrodami! I pizza! UWAGA: WEJŚCIE OD STRONY KAWIARNI!

Plan wydarzenia: 18:00-18:10 Wstęp 18:10-18:25 SPOCK - Jan Bassa: “Uczenie ze wzmocnieniem z biblioteką Stable Baselines” 18:25-18-40 SPOCK - Oskar Walawender: “Modele multimodalne: Integracja tekstu, obrazu i dźwięku” 18:40-18:55 SPOCK - Mateusz Szerwiński: “Wojna Tytanów: PyTorch vs. Tensorflow” 18:55-19:10 Przerwa 19:10-19:55 Bartosz Sokorski: “Jestem na to za leniwy - dbanie o jakość kodu w Pythonie” 19:55-20:05 Konkurs

O prelegentach: Jan Bassa Interesuję się tematyką sztucznej inteligencji i wdrażaniu jej w praktycznych przykładach. Prywatnie interesuję się antropologią, podróżami i literaturą.

Oskar Walawender Jestem studentem informatyki i członkiem koła SOMSIAD, gdzie rozwijam pasje związane ze sztuczną inteligencją i analizą danych. Aktywnie uczestniczę w warsztatach i projektach, zdobywając praktyczne umiejętności w pracy z nowoczesnymi technologiami.

Mateusz Szerwiński Student inżynierii biomedycznej. Poza studiami, uwielbia bieganie i górskie wędrówki.

Bartosz Sokorski Programista z zawodu i pasji. Po godzinach utrzymuje kilka projektów open-source. Głównie zainteresowany rozwojem aplikacji CLI. Zawodowo związany z STX Next.

O wykładach: Jan Bassa: “Uczenie ze wzmocnieniem z biblioteką Stable Baselines” Podczas prelekcji opowiem o Reinforcement Learning (RL) z użyciem biblioteki Stable Baselines3. Omówię, jakie wyzwania stoją przed wdrażaniem RL w praktyce, a następnie pokażę, jak Stable Baselines upraszcza cały proces. Przedstawię praktyczny przykład treningu agenta w popularnym środowisku CartPole oraz omówię wyniki wizualizowane za pomocą TensorBoard. Wspomnę o zmianach w Stable Baselines z wersji na wersję.

Oskar Walawender: “Modele multimodalne: Integracja tekstu, obrazu i dźwięku” W erze ogromnych zbiorów danych modele multimodalne otwierają drzwi do inteligentniejszego i bardziej złożonego przetwarzania informacji. Podczas prezentacji odkryjesz, jak technologie takie jak CLIP i Flamingo łączą tekst, obraz i dźwięk, aby sprostać współczesnym wyzwaniom.

Mateusz Szerwiński: “Wojna Tytanów: PyTorch vs. Tensorflow” W tej prezentacji porównane zostaną dwa wiodące frameworki w dziedzinie uczenia maszynowego: PyTorch i TensorFlow. Omówione zostaną kluczowe aspekty obu rozwiązań, a także wskazówki dotyczące wyboru między nimi — dla kogo lepszym wyborem będzie TensorFlow, a dla kogo PyTorch. Przedstawione zostaną również perspektywy rozwoju tych narzędzi w przyszłości.

Bartosz Sokorski: “Jestem na to za leniwy - dbanie o jakość kodu w Pythonie” Dbanie o jakość i czytelność kodu bywa żmudne i nużące, a często potrafi wywołać wiele dyskusji w zespole. W swojej prezentacji chcę pokazać jak ułatwić sobie pracę i dbać o jakość kodu z minimalnym wysiłkiem.

O konkursach: Rozdamy masę nagród! W naszym tradycyjnym kahoocie do wygrania będzie:

• 1 x licencja na dowolny produkt JetBrains!

Spośród osób zapisanych na nasze spotkanie na meetup.com wylosujemy zwycięzce, który otrzyma:

• 1 x licencja na dowolny produkt JetBrains!

Uwaga: losowanie będą mogły wygrać tylko osoby, które zapiszą się na nasze spotkanie na meetup.com. Zależy nam na tym dlatego, że od liczby zapisanych będzie zależała liczba licencji, jakie będziemy w stanie zdobyć dla Was w przyszłości.

O sali: Tradycyjnie spotykamy się w bydgoskich Młynach Rothera (Spichrz Zbożowy, 2. piętro). Uwaga: wejście od strony kawiarni! Również tradycyjnie zapraszamy na pizzę! I, też tradycyjnie, zapraszamy na afterparty, tym razem w Cybermachinie.

Partnerzy Młyny Rothera Serdecznie zapraszamy!

PyTorch isn't just another framework; it's the de-facto standard for Deep Learning. Our community is dedicated to bringing together PyTorch users in London and those with a profound interest in ML and AI. This is your platform to share experiences, network, seek advice, and initiate collaborations.

This event is packed with opportunities to learn, connect, and exchange ideas and knowledge with experts from both Revolut and the data industry.

Join us for exclusive talks from Revolut speakers and our partner, PyTorch, and dive into the world of Data Science. Here’s what you can look forward to at the meet-up.

We'll have 3 amazing speakers at the event: Vadim Andronov, Deep Learning Engineer (Computer Vision) at Revolut Nikolay Falaleev, Head of AI at Sportlight Technology Vincent Moens, Applied Machine Learning Research Scientist at Meta Event Details: Date: Thursday, 21st November 2024 Time: 18:00 (GMT+4) Where: Revolut HQ Office Agenda:

• 17:30–18:00 \| Gathering
• 18:00–18:15 \| Welcome & Intro
• 18:15–18:55 \| Presentation: "End-to-end Visual Document Understanding" by Vadim Andronov + Q&A session
• 18:55–19:00 \| Refreshment Break
• 19:00–19:40 \| Presentation: "Optimising Video Pipelines for Neural Network Training with PyTorch" by Nikolay Falaleev + Q&A session
• 19:40–19:50 \| Refreshment Break
• 19:50–20:30 \| Presentation: "Breaking Free from CPU Bottlenecks: Unlocking PyTorch's Potential for Tiny Models" by Vincent Moens + Q&A session
• 20:30–21:30 \| Networking

Registration:

Fill out this form to receive your exclusive invitation!

Please note that this is a private event, and you'll need an invitation to attend. Since spots are limited. You’ll receive your invitation by email 3 days before the event. If the event is fully booked, you will be placed on a waiting list and notified if a spot becomes available.

Important: Only those who have submitted the registration form will be considered for an invitation. For security purposes, only invited guests on the list will be granted access to the premises; anyone not on the list will not be allowed entry.

Join us this September 25th for the London PyTorch Meetup #19

About the Event: PyTorch isn't just another framework; it's the de-facto standard for Deep Learning. Our community is dedicated to bringing together PyTorch users in London and those with a profound interest in ML and AI. This is your platform to share experiences, network, seek advice, and initiate collaborations. We've curated an exceptional line-up of speakers who will share their expertise on PyTorch applications across diverse sectors. And yes, we're on the brink of resuming our much-missed in-person events!

Our Speakers:

Nikolay Falaleev (Head of AI at Sportlight) "Optimising Video Pipelines for Neural Network Training with PyTorch"

Vincent Moens, Research Engineer at Meta (core PyTorch team) "Breaking Free from CPU Bottlenecks: Unlocking PyTorch's Potential for Tiny Models"

Looking forward to an engaging and informative session with all of you!

It is time for another edition of the PyData Eindhoven meetup. Join us on the 5th of September at Vanderlande for an evening of technical talks for technical people. We are organising this event around the theme: ML Automation in the logistics domain.

17:30 - Doors Open 18:00 - Welcome 18:05 - Welcome from Vanderlande 18:10 - Secure MLOps with MLFlow simplifying integration tests at NS by Luuk van der Velden (Codebeez) 18:40 - Dinner 19:15 - Practical use of MLOps tools for automated visual inspection by Thomas Woudsma (Prodrive Technologies) 19:45 - Collaborative picking guided by Deep Reinforcement Learning by Kasper Hendriks (Vanderlande) and Igor Smit (TU/e) 20:15 - Wrap/up20:20 - Networking Drinks --- Title: Secure MLOps with MLFlow simplifying integration tests at NS Description: Cloud security teams want us to duplicate our resources across subscriptions and environments for security and privacy reasons. Data science applications struggle to be tested in the absence of real-world production data. At NS we performed integration tests on non-production data, but this broke down often and it was costly to maintain two separate data lakes (staging, production). Here I present our open source MLFlow client (Python) that allowed us to reduce the complexity of our test setup and reduce maintenance costs. Presenter: Luuk van der Velden Company: https://codebeez.nl LinkedIn: https://www.linkedin.com/in/ljjvandervelden/ --- Title: Practical use of MLOps tools for automated visual inspection Description: Using deep learning to solve many real-world problems is becoming more and more ubiquitous. Especially for applications using image data, models created with TensorFlow or PyTorch achieve state-of-the-art performance. The next challenge is to keep these models at peak performance in production scenarios. Like with DevOps for software development, MLOps practices like data engineering, model engineering, and model serving and monitoring have been defined to streamline ML applications.

I this talk we will look at the application automated visual inspection in the electronics manufacturing facilities of Prodrive Technologies. Some of the latest model architectures are shown together with MLOps tools like MLflow (experiment tracking and model registration) and Seldon Core (model deployment and serving) including some examples in Python to give some technical insight into how these tools can be used. Presenter: Thomas Woudsma Company: https://prodrive-technologies.com LinkedIn: https://www.linkedin.com/in/thomaswoudsma/ --- Title: Collaborative picking guided by Deep Reinforcement Learning Description: Collaborative case picking is Vanderlande's next step in improving picking efficiency in warehousing. This is achieved by leveraging the collaboration between human pickers and autonomous vehicles. However, the complexity and inherited uncertainty of warehouse operations make orchestrating this collaboration challenging for existing warehouse management software. In this talk, we will look at how we can model collaborative case picking effectively using Python and showcase how this enables us to use DRL to improve over traditional optimization approaches. Presenter: Kasper Hendriks and Igor Smit. Company: Vanderlande LinkedIn: https://www.linkedin.com/in/kasper-hendriks/ & https://www.linkedin.com/in/igor-smit/

--- Attendee profile: * People who want to learn about Python, Julia, R in the context of Data * Interested in seeing how others use Python for solving real problems * Interested in being part of the Python Community in Eindhoven * No Recruiters * No Sales * Only people who speak Data

TIMESERIES #FORECASTING #LLMS #LANGCHAIN

In person in Heidelberg and live on PyData.TV on YouTube

Agenda

18:00 Doors open

18:30 Welcome

18:45 Beyond Toy Datasets: Timeseries Forecasting for Real Business Problems - Robert Haase (AI Scientist @ paretos)

19:15 Break: Networking with snacks and beverages

20:00 How to Leverage the Full Potential of LLMs for Your Business with Langchain - Leon Ruddat (AI Research Specialist @ SNOCKS)

20:30 Lightning Talks

20:45 Networking with snacks and beverages

21:30 End

Lightning Talks

Join us by contributing a five-minute lightning talk!

Fill out this form.

How to sign up

It's important for us to make this meet up happen in a responsible way. We have limited seats available only.

How to join remotely

Join the live stream on YouTube.

Q&A

Ask via Slido

This event will be in English.

----

About this meetup:

Talk #1

Robert Haase (AI Scientist @ paretos)

Beyond Toy Datasets: Timeseries Forecasting for Real Business Problems - Common Pitfalls and (Some) Solutions

Real-life timeseries datasets e.g. from retailer or logistic companies often come with high complexity such as a high level of sparsity and heterogenity as well as most of the times it is highly imbalanced and driven by external factors. In this talk Robert will give an overview of what are the common pitfalls when working with real-life data in a timeseries context and provide a set of possible solutions to turn these projects into a success-story.

Robert earned both his Bachelor's and Master's degrees in Physics from the University of Heidelberg, specializing in Condensed Matter Physics and Computational Physics. During his Master's thesis in 2020, he advanced existing NLP Transformer architectures for timeseries applications. This involved Robert working extensively with uncertainty quantifications and normalizing flows. Since the beginning of 2021, he has been employed at Paretos, where the primary focus of his work lies in Timeseries Forecasting, specifically demand forecasting. Robert has a keen interest in combining traditional statistical methods with deep learning techniques.

Talk #2

Leon Ruddat (AI Research Specialist @ SNOCKS)

How to Leverage the Full Potential of LLMs for Your Business with Langchain

In this talk, Leon will introduce Langchain, a framework for developing applications powered by language models. He'll explore how to work with it, which use cases are most suited for this framework, and most importantly, which features of Langchain can deliver the greatest value from his perspective.

Leon lives in Heidelberg. He holds a Master's degree in Mechatronics with a focus on Artificial Intelligence from Hochschule Mannheim. During his studies there, he was employed as an AI Scientist for almost three years, with his research mainly focusing on autonomous driving. Since April, he has been working at Snocks as an AI Scientist. There, he delves into the exciting world of Large Language Models (LLMs) and explores their applications in e-commerce. He currently leads a team of three developers with whom he collaborates on innovative AI solutions. His expertise lies particularly in working with PyTorch, CV-based AI frameworks, Langchain, and well-known LLMs such as GPT. What drives him? The "Why-Not-Spirit". His motto: "Just do it and build awesome AI tools."

----

Lightning Talks:

1. Locating the spiral arms of the Milky Way using t-sne - Dr. Bertrand Lemasle

2. Developing an AI-based Business Data Analyst using OpenAI Function Calling - Bernhard Schäfer

3. Open-Source Science (OSSci) - Tim Bonnemann

Acknowledgements

Also a big thank you to our sponsors:

• NUMFOCS, for promoting open source software.
• KÖNIGSWEG, for supporting the organization.
• paretos, for hosting the meetup.

Contact

If you have any questions or suggestions, please feel free to contact us via:

• Meetup
• Want to speak? Submit a talk here.
• Interested in hosting an event? Here's our Info-Deck & contact to the organisers!