Der Swiss Python Summit ist eine zweitägige Konferenz rund um die Programmiersprache Python. Neben einem allgemeinen Python-Tag zur Vielfalt der Anwendungsmöglichkeiten bietet das Programm einen eigenen Data Science Day. Das Event lädt zum Vernetzen mit der Community, zum Erfahrungsaustausch und zum Entdecken aktueller Trends ein.
Wir freuen uns sehr, dass unsere Kollegen Dominik Traxl mit seinem Vortrag No more model training? TabPFN v3 vs. tuned trees, Jonas Böer mit seinem Vortrag Pulumi: Making Infrastructure as Code fun again with Python und Florian Wilhelm mit seinem Vortrag Sentinel Values in Python: Why None Is Not Enough vertreten sein werden.
Abstract von Dominik Traxls Vortrag No more model training? TabPFN v3 vs. tuned trees:
What if you never had to train or tune a model again? That is the pitch behind
tabular foundation models: state-of-the-art predictions with no training, no
hyperparameter search, no feature engineering. Just call .fit() and .predict().
The headliner, TabPFN, whose earlier version was published in Nature, now adds
claims aimed straight at the gradient-boosted trees most of us reach for by
default. TabPFN v3 promises to ingest raw text natively, scale toward a million
rows, support multi-class classification natively and offer a
„thinking mode“ for extra quality. Too good to be true? I put it on trial.
This talk walks through a systematic, reproducible Python benchmark that puts
TabPFN, exactly as it ships, up against a best-effort, Optuna-tuned pipeline of
XGBoost, LightGBM, and CatBoost on the messy public Steam Dataset 2025:
numerics, booleans, categoricals, date-times, free-text, embeddings, skewed
distributions, NaNs. The case spans three tasks: binary sentiment classification,
multi-class publisher prediction, and Metacritic-score regression.
We stress-test each vendor claim one at a time: raw text versus sentence
embeddings versus no text, out-of-the-box versus cleaned data, default versus
thinking mode, small samples up to large, few to many classes.
But the real takeaway is the method: how to benchmark a hyped claim fairly, why
the metric you pick can quietly decide the winner, and the data-leakage traps you
must defuse before you trust a single result.
The honest answer to „Too good to be true?“ is: „it depends“. And knowing what
it depends on is the point. You’ll leave with a feel for when a tabular
foundation model earns its keep, when a tuned tree still wins, and how to put
the next „no training required“ promise on trial yourself.
Abstract von Jonas Böers Vortrag Pulumi: Making Infrastructure as Code fun again with Python:
We don’t need to use Python only for developing applications and pipelines, we can also use it to maintain our infrastructure. Pulumi allows us to describe our infrastructure as a Python program. As this gives us access to Python’s flexibility, readability, power, and joy, it has quickly become my favourite way to manage infrastructure, no matter if for a small team’s ML application or for a multinational company’s analytics platform. Let me share this joy with you!
In this talk I will show you how Pulumi works and what it does better than other IaC tools: Among others, I will talk about custom resources and resource grouping, state versioning and understanding your infrastructure deployment as a program to be run.
Caution: This talk may lead to strong feelings of „why haven’t we always done it like this?“
Abstract Florian Wilhelms Vortrag Sentinel Values in Python: Why None Is Not Enough:
Every Python developer has used None as a default but what happens when None is also a valid value? This talk explores sentinel values as a first-class concept in Python. We’ll uncover why None is semantically overloaded, how built-in sentinels like NotImplemented and ... (Ellipsis) quietly power Python’s object model, and what happens when you try to create your own. The central challenge: Python’s type system has no good answer for sentinel values. We’ll walk through why Optional, Literal, and custom classes all fall short, look at real-world workarounds including Pydantic’s experimental MISSING, and discuss what PEP 661’s deferral means for your code today. You’ll leave with a clear mental model and practical guidance for handling absence in typed Python.