PyCon DE & PyData 2025
2025-04-23..2025-04-25
Darmstadt, Germany
as the organisers put it
Join PyCon DE & PyData 2025 in Darmstadt (Frankfurt), April 23-25! Germany’s largest Python and Data Science conference with talks, workshops, and community events like DjangoGirls and sprints. Be part of the action!
| title | author | length |
|---|---|---|
| Conformal Prediction: uncertainty quantification to humanise models | Vincenzo Ventriglia | 30m |
Conformal prediction turns point outputs into prediction sets with mathematical coverage by computing nonconformity scores on a reserved calibration set and… | ||
| Taking Control of LLM Outputs: An Introductory Journey into Logits | Emek Gözlüklü | 30m |
Shows how raw logits from LLAMA-3B-instruct drive deterministic token scores and how direct logit manipulation—token masking via constraint decoding—forces… | ||
| Benchmarking Time Series Foundation Models with sktime | Benedikt Heidrich | 27m |
sktime exposes multiple time series foundation models (Morai, Tiny Times Mixer, Kronos, TimesFM, TimeMOE, MomentFM, etc.) behind a unified API and a… | ||
| expectation: A modern take on statistical A/B testing with e-values and martingales | Jako Rostami | 47m |
A new Python library called expectation implements e-values and martingales for sequential A/B testing, change‑point detection and confidence sequences, with… | ||
| PyData Stack: Pure Python open source data platforms | Eric Thanenthiran | 40m |
Builds a pure‑Python data platform that actually runs locally and handles batch workloads up to tens of gigabytes using DLT, DuckDB, dbt, Dagster, Streamlit… | ||
| How to use Data Science Superpowers in real life, a Bayesian perspective | Tim Lenzen | 29m |
Demonstrates applying Bayesian update with PyMC to decide when to take an Azure certification: starts with a flat Beta prior that estimates >60% chance of… | ||
| LLM Inference Arithmetics: the Theory behind Model Serving | Luca Baggi | 30m |
Argues that transformer inference is inherently serial while training is parallel, so training needs ~3× the FLOPs of inference and attention compute scales… | ||
| Mastering Demand Forecasting: Lessons from Europe's Largest Retailer | Moreno Schlageter, Yovli Duvshani | 28m |
Retail demand forecasting runs on a feature‑based GBM pipeline using LightGBM, training one model per item‑warehouse and running hundreds of thousands of… | ||
| The aesthetics of AI: from cyberpunk to fascism | Laura Summers | 32m |
Argument ties brutalist and cyberpunk visuals to power, showing concrete examples: Clippy is offline and playful while Claude is cloud‑hosted and… | ||
| Conquering PDFs: document understanding beyond plain text | Ines Montani | 31m |
A practical wrapper integrates Dockling with spaCy to extract text plus layout spans and bounding boxes from PDFs, feed that Dockling document into existing… | ||
| Building Reliable AI Agents for Publishing: A DSPy-Based Quality Assurance Framework | Simonas Černiauskas | 30m |
A newsroom-grade pipeline built on DSPy composes modular text transformation graphs to run multistep generation (chain-of-thought), automatic knowledge-graph… | ||
| Secure "Human in the Loop" Interactions for AI Agents | Juan Cruz Martinez | 27m |
Argues that human oversight must sit between planning and execution for stateful AI agents and shows a working pattern using LangChain interrupts, persisted… | ||
| Reinforcement Learning Without a PhD: A Python Developer’s Journey | Jochen Luithardt | 29m |
Claims a practical path: a supervised simulator learned from historical campaign data predicts next states so agents train with far less data, an ensemble… | ||
| Getting Started with Bayes in Engineering: Implementing Kalman Filters with RxInfer.jl | Victor Flores Terrazas | 31m |
Shows a concrete implementation of state-space Bayesian estimation using RxInfer.jl, building A/B/C matrices from structural mass/damping/stiffness, and… | ||
| Open Table Formats in the Wild: From Parquet to Delta Lake and Back | Franz Wöllert | 46m |
Attempts to replace a Parquet bronze layer with Delta Lake hit three concrete surprises: a 1.3 TB compaction test shows liquid clustering cuts API calls and… | ||
| Power up your Polars code with Polars extention | Cheuk Ting Ho | 01h 16m |
Shows how to replace Python apply calls by writing Polars extensions in Rust using PyO3 and Maturin to register custom Polars expressions, implement… | ||
| Beyond FOMO — Keeping Up-to-Date in AI | Carsten Frommhold | 29m |
Rapid AI change costs focus and well being, so prioritize concrete habits: run a team knowledge culture with no stupid questions and a two-step… | ||
| From Tensors to Clouds — A Practical Guide to Zarr V3 and Zarr-Python 3 | Sanket Verma | 29m |
Zarr V3 and Zarr-Python 3 formalize interoperability and extensibility: a new abstract Store ABC lets users register custom stores (examples include git-like… | ||
| supplyseer: Computational Supply Chain with Python | Jako Rostami | 01h 18m |
A Python library called supplyseer provides runnable modules: a demand simulator using geometric Brownian motion to generate 1,000s of sample paths over a… | ||
| Autonomous Browsing using Large Action Models | Arne Grobrügge, Nico Kreiling | 30m |
A working autonomous browser agent stacks a Lavage-based multiagent pipeline with a Selenium grid-backed browser, a multimodal supervision engine that reads… | ||
| Is Prompt Engineering Dead? How Auto-Optimization is Changing the Game | Oleh Kostromin, Iryna Kondrashchenko | 29m |
Automatic prompt optimization treats prompts as trainable parameters and demonstrates that methods like bootstrap few-shot, MIPRO, and TextGrad can replace… | ||
| Beyond Basic Prompting: Supercharging Open Source LLMs with LMQL's Structured Generation | Christiaan Swart | 46m |
LMQL enables interweaving prompts and Python for LMQL‑style structured generation, using logit biasing, type constraints and control flow to produce validated… | ||
| Build a personalized Commute agent in Python with Hopsworks, LangGraph and LLM Function Calling | Javier de la Rúa Martínez | 29m |
A working Python pipeline combines Hopsworks, LangGraph and LLM function calling to power a personalized commute agent that ingests Stockholm public‑transport… | ||
| PDFs - When a thousand words are worth more than a picture (or table). | Caio Benatti Moretti | 31m |
Claims a vision-driven pipeline that uses multimodal models to turn PDF visuals into descriptive text chunks, improving retrieval vs raw parsing or… | ||
| From Idea to Integration: An Intro to the Model Context Protocol (MCP) | Julian Beck | 25m |
Model Context Protocol (MCP) presents a concrete standard for connecting LLMs to tools and services and demonstrates a working SQLite conference-schedule… | ||
| Building Serverless Python AI skills as WASM components | Moritz Althaus | 35m |
Claims that combining WebAssembly component model with serverless principles lets teams ship Python AI skills as WASM components that the runtime runs and… | ||
| Modern NLP for Proactive Harmful Content Moderation | Daryna Dementieva | 26m |
Compares GPT‑4, Llama/Mistral and other LLMs to encoder transformers (mT5, mT0, mBART), finding out‑of‑the‑box LLM classification often performs poorly while… | ||
| From Trees to Transformers: Our Journey Towards Deep Learning for Ranking | Theodore Meynard, Mihail Douhaniaris | 29m |
GetYourGuide replaces an XGBoost ranker with a production PyTorch transformer over nine months, keeping throughput and latency acceptable after tuning CPU… | ||
| Jeannie: An Agentic Field Worker Assistant | Andrei Beliankou, Jose Moreno Ortega | 25m |
Jeannie is an agentic field assistant that combines task preparation PDF generation via Playwright with SQLAlchemy database calls, external weather APIs… | ||
| Generative-AI: Usecase-Specific Evaluation of LLM-powered Applications | Homa Ansari | 29m |
Presents a practical framework that evaluates LLM applications on quality, price, and speed by building use-case benchmark sets with subject-matter experts… | ||
| Reasonable AI | Kristian Kersting | 42m |
Argues that scaling alone fails and proposes a modular “reasonable AI” stack: open bilingual LLMs with continuous pretraining on German, a memory-saving… | ||
| The earth is no longer flat - introducing support for spherical geometries in Spherely and GeoPandas | Joris Van den Bossche | 36m |
Claims a working Python spherical-geometry engine: Spherely v0.1 wraps Google's S2 via pybind11 and provides area, distance and basic predicates with… | ||
| Streamlining the Cosmos: Pythonic Workflow Management for Astronomical Analysis | Raphael Hviding | 35m |
Astronomical pipelines scale from single-galaxy processing to orchestration across HPC clusters, using Snakemake to auto-generate job submissions (no… | ||
| Are LLMs the answer to all our problems? | Dr. Maria Börner | 28m |
Claims that LLMs solve everything meet concrete limits: visual outputs show gender and racial bias in training data (DALL·E example), GDPR means personal data… | ||
| Building an Open Source RAG System for the United Nations Negotiations on Global Plastic Pollution | Rahkakavee Baskaran, Teresa Kroesen, Anna-Lisa Wirth | 34m |
An open-source RAG system supports the UN plastic-pollution treaty by ingesting chunked documents with multilingual sentence-transformer embeddings, storing… | ||
| Beyond Agents: What AI Strategy Really Needs in 2025 | Alexander CS Hendorf | 47m |
GTC observations condense into concrete shifts: NVIDIA markets a compact developer box with 128 GB of VRAM and about one teraflop for local model work… | ||
| Guiding data minds: how mentoring transforms careers for both sides | Anastasia Karavdina | 31m |
Data from a personal mentoring survey shows typical engagements drop quickly: median churn sits around two months and about half stop by three months, while a… | ||
| Citation is Collaboration: Software Recognition in Research and Industry | Ivelina Momcheva | 29m |
Analysis of 2011–2024 astronomy publications shows 93% of astronomers use Python while formal mentions peak at only about 10–12% of papers and most appear in… | ||
| Zero Code Change Acceleration: familiar interfaces and high performance | Tim Head | 24m |
Demonstrates keeping standard pandas/scikit-learn code while switching implementations under the hood: an import-hook that replaces modules with wrappers that… | ||
| Driving Trust and Addressing Ethical Challenges in Transportation through Explainable AI | Natalie Beyer | 31m |
Case study with Init in Karlsruhe builds a real-time dispatch support system using lightGBM trained on Init’s operations logs, surfacing per-prediction… | ||
| Lessons learned in bringing a RAG chatbot with access to 50k+ diverse documents to production | Bernhard Schäfer, Nico Mohr | 45m |
Production RAG system at Merck Electronics indexes 50,000+ documents from SharePoint and experiment writeup DBs, built on Palantir Foundry with PySpark/Polars… | ||
| Analyze data easily with duckdb - and the implications on data architectures | Matthias Niehoff | 28m |
DuckDB is an in-process analytical engine that queries Parquet/JSON over HTTP and S3 (example: reads a 2.5 GB JSON blob) and can attach tiny catalog files (few… | ||
| Composable AI: Building Next-Gen AI Agents with MCP | Martin Seeler | 30m |
Explains how the Model Context Protocol (MCP) turns APIs into reusable tool servers so agents call tools (via function calling) instead of custom integrations… | ||
| Dataframely — A declarative, 🐻❄️-native data frame validation library | Daniel Elsner, Oliver Borchert | 30m |
Dataframely is a Polars-native validation library that ships declarative schemas, column checks and cross-column constraints expressed as Polars expressions… | ||
| BayBE: A Bayesian Back End for Experimental Planning in the Low-To-No-Data Regime | Martin Fitzner, Alexander Hopp, Adrian Šošić | 01h 30m |
BayBE is an open‑source Apache‑2 experimental planner that combines Bayesian optimization with BoTorch/GP tools, a probabilistic surrogate, and acquisition… | ||
| Accuracy Is Not Enough: Building Trustworthy AI with Conformal Prediction | Chris Aivazidis | 44m |
Conformal prediction converts point outputs into prediction sets with a user‑specified coverage (examples use 95% coverage) under the exchangeability… | ||
| Machine Learning Models in a Dynamic Environment | Isabel Drost-Fromm | 41m |
Users adapt to models and break assumptions: historical examples (SEO gaming, email spam) show adversarial cat‑and‑mouse dynamics, LLMs trained on the web… | ||
| Cache me if you can: Boosted application performance with Redis and client-side caching | David Maier | 32m |
A concrete recipe for client-side caching with Redis: use redis-py 5.1.1+ with RESP3 + Redis 7.4+, enable the client cache (LRU by item count, default key =… | ||
| The future of AI training is federated | Chong Shen Ng | 01h 22m |
Argues that production AI training moves to federated setups and shows a working Flower stack: flower new, flower run, flower list, flower log, flower stop… | ||
| Scalable Python and SQL Data Engineering without Migraines | Dirk Jung | 32m |
Live demo builds a Snowflake pipeline for a franchise food‑truck case: create a database and isolated virtual warehouse, stage S3 files, load ~20 GB in the… | ||
| Securing RAG Pipelines with Fine Grained Authorization | Sohan Maheshwar | 33m |
Shows a concrete ReBAC implementation using SpiceDB with Pinecone embeddings, OpenAI for the LLM and LangChain to orchestrate; demonstrates a pre-filter flow… | ||
| Chasing the Dark Universe with Euclid and Python: Unveiling the Secrets of the Cosmos | Guadalupe Canas Herrera | 46m |
ESA Euclid maps billions of galaxies from L2 to probe dark matter and dark energy, launched July 2023 and already releasing 63 sq deg of imaging; a public… | ||
| Distributed file-systems made easy with Python's fsspec | Einat Orr, Barak Amar | 38m |
fsspec provides a single Python API for local and cloud stores, showing concrete code: pandas reads a CSV/Parquet via fsspec with S3/GCS backends, parquet… | ||
| Machine Reasoning and System 2 Thinking | Andy Kitchen | 31m |
Shows that recent models jump from ~3–4% on the ARC challenge to much higher accuracy when reasoning is unified: tools+agency+induction+transduction inside a… | ||
| Bias Meets Bayes: A Bayesian Perspective on Improving Model Fairness | Vince Nelidov | 31m |
Simulation uses a 90%/10% split with 450 majority and 50 minority samples and a data generator whose true risk rate is 36%. A vanilla logistic model (no group)… | ||
| Transformers for Game Log Data | Fabian Hadiji | 45m |
Applies a GPT-style decoder to game telemetry using PyTorch and nanoGPT-inspired code, trains on the Blade and Soul dataset (175M events, 77 base event types)… | ||
| Scraping LEGO for Fun: A Hacky Dive into Dynamic Data Extraction | Peter Lodri | 38m |
Shows an LLM-driven scraper that replaces fragile XPath/CSS logic with prompted visual element selection using Chromium hooked to GPT‑4o and frameworks like… | ||
| Decoding Topics: A Comparative Analysis of Python’s Leading Topic Modeling Libraries Using Climate C | Lisa Andreevna Chalaguine | 45m |
Compares LDA, Top2vec, and BERTopic on 14,000 climate policy text blocks and four industry tasks: similarity, classification, real-time insertion, and dynamic… | ||
| Blazing-Fast Python in Your Database: Unlocking Data Science at Scale with Exasol | Alexander Stigsen | 29m |
Exasol runs native Python UDFs inside an MPP database using script language containers, a bucket file system for distributing models, and SQL orchestration to… | ||
| AI in Reality Fireside Chat: Enterprise AI & Open‑Source Innovation | Ines Montani, Alexander CS Hendorf, Alexander Beck, Walid Mehanna | 58m |
Covers concrete enterprise moves: Merck runs a four‑year AI program for 63,000 employees, prototypes LLMs, then abandons an internal clone for a partnership to… | ||
| Unlocking the Predictive Power of Relational Data with Automated Feature Engineering | Alexander Uhlig | 01h 25m |
Uses the H&M Fashion dataset to auto‑generate tabular features for churn classification and item‑sales regression, applying getML FastProp propositionalization… | ||
| Multi-tenant Conversational Analytics | Rodel van Rooijen | 30m |
Built a multi-tenant conversational analytics system for cash-management customers that actually queries BigQuery with row-level security, stores embeddings in… | ||
| PyLadies Panel: AI Skills & Careers | Anastasia Karavdina, Jesper Dramsch, Guadalupe Canas Herrera, Tereza Iofciu | 01h 02m |
Generative AI is speeding real work: large models enable processing of massive astronomy and weather datasets but require rigorous validation to avoid false… | ||
| AI Agents of Change: Creating, Reflecting, and Monetizing | Paloma Andrade de Oliveira, Tereza Iofciu | 01h 26m |
Workshop shows how to build an agent on the OpenServe platform with a Python SDK wrapper around a TypeScript SDK, using ngrok for a local tunnel to expose a… | ||
| Unforgettable, that's what you are: Evaluating Machine Unlearning and Forgetting | Katharine Jarmul | 30m |
Shows that large models memorize both common and novel (outlier) examples and that attackers can use shadow models + discriminators for membership inference… | ||
| A11y Need Is Love (But Accessible Docs Help Too) | Smera Goel | 43m |
Argues that accessibility belongs in the foundation of docs not as a cherry on top, using the PyData Sphinx Theme audit and fixes as proof: runs Microsoft… | ||
| Career Path Experience Stories | Kristina Khvatova | 01h 48m |
Panelists trace concrete pivots: one built an ERP from scratch to process paper sales reports for a record label and taught themself programming; another ships… | ||
| Building versatile operating setups for real world use and testing with Python and the Raspberry Pi | Jens Nie | 30m |
A modular control platform uses Raspberry Pi Compute Module hardware (moving from CM3B+ toward CM5) and Python apps per component communicating via Redis… | ||
| The Foundation Model Revolution for Tabular Data | Noah Hollmann, Frank Hutter | 29m |
Claims state-of-the-art on small tabular tasks, beating XGBoost in 96% of use cases and reaching similar accuracy with 50% of the training data. Model ingests… | ||
| Reinforcement Learning for Finance | Yves J. Hilpisch | 01h 31m |
Covers applying Deep Q-Learning to finance with runnable notebooks (TensorFlow originals plus PyTorch notebooks and Google Colab links), trading and hedging… | ||
| Building a HybridRAG Document Question-Answering System | Darya Petrashka | 30m |
Combining vector RAG retrieval with an explicit knowledge graph produces a hybrid system that raises correct-answer rates in a law/finance benchmark from ~57%… | ||
| How Narwhals is silently bringing pandas, Polars, DuckDB, PyArrow, and more together | Marco Edward Gorelli | 29m |
Narwhals provides a lightweight compatibility layer that lets libraries use pandas, Polars, PyArrow, DuckDB and more without adding dependencies, claiming zero… | ||
| Pipeline-level differentiable programming for the real world | Alessandro Angioi | 25m |
Presents an open-source system that wraps Python modules as Tessract Core containers exposing a containerized REST API and auto-generates Pydantic schemas… | ||
| Beyond DALL-E: Advanced Image Generation Workflows with ComfyUI | René Fa | 28m |
ComfyUI workflows combine Stable Diffusion XL, LoRA and plugins to achieve precise results: demonstration pipelines run on runpod.io with Docker and a GitHub… | ||
| Extending Python with Rust, Mojo, Cuda and C and building packages | Wolf Vollprecht, Ruben Arts | 29m |
Pixie is a Rust-written package manager that installs conda and pip packages, compilers and CUDA toolchains as a single binary and produces precise pixie.lo… | ||
| High-performance dataframe-agnostic GLMs with glum | Martin Stancsics | 29m |
glum is a Python-first GLM implementation with a C++ TabMat backend and Cython hot paths that claims up to ~1,000x speedups on categorical sandwich-product… | ||
| Agentic AI: Build a Multi-Agent Application with CrewAI | Alessandro Romano | 01h 04m |
Live demo builds a working multi agent app with CrewAI showing code, not slides, and four hands on assignments. It creates a two agent pipeline that takes a… | ||
| Topological data analysis: How to quantify "holes" in your data and why? | Ondrej Draganov | 43m |
Pipeline uses persistent homology to turn point clouds into fingerprints: grow discs, record births/deaths, produce persistence diagram and vectorize with a… | ||
| Is your LLM any good at writing? Benchmarking on creative writing and editing tasks | Azamat Omuraliev | 29m |
A systematic benchmark tests 18 models on 15 concrete writing tasks spanning date & greeting formatting, following and avoiding explicit style instructions… | ||
| Intuitive A/B Test Evaluations for Coders | Thomas Mayer | 29m |
Argues that resampling gives coders an intuitive path to A/B decisions by coding rather than advanced stats, using numpy random and SciPy to reproduce results… | ||
| From stockouts to happy customers: Proven solutions for time series forecasting in retail | Robert Haase | 31m |
Retail forecasting solutions center on concrete fixes: handle censored demand with seasonal imputation or by passing count out-of-stock days as features and… | ||
| Using Causal thinking to make Media Mix Modeling | Carlos Trujillo | 27m |
Shows a working pipeline that builds Bayesian structural media-mix models with PyMC and graph definitions in PyTensor, modeling adstock, saturation and… | ||
| Offline Disaster Relief Coordination with OpenStreetMap and FastAPI | Jannis Lübbe | 32m |
An offline map server running on a Raspberry Pi Zero 2W serves Germany-wide OpenStreetMap vector tiles (4.2 GB PBF → MBTiles) via FastAPI + MBTiles to a local… | ||
| Using Python to enter the world of Microcontrollers | Jens Nie | 34m |
Demonstrates running real Python runtimes on modern chips and actually decoding household meters: MicroPython on Pico W and ESP32 handle parsing D0 and SML… | ||
| 3 Ways to Speed up Your Regression Modeling in Python | Alexander Fischer | 33m |
Benchmarks show a 20 million–row regression with 60 one‑hot features runs ~10s with a naive normal‑equation, speeds ~30% using sparse matrices (scikit‑learn /… | ||
| What's inside the box? Building a deep learning framework from scratch. | Oleh Kostromin | 01h 22m |
Builds a minimalist deep learning framework in Python in under 90 minutes that uses a NumPy backend, implements elementwise ops, matrix multiply, ReLU… | ||
| Enhancing RAG with Fast GraphRAG and InstructLab: A Scalable, Interpretable, and Efficient Framework | Tuhin Kanti Sharma | 31m |
Systematically compares plain RAG, Microsoft GraphRAG, HippoRAG and a new FastGraphRAG using a 100 Q&A gold set generated by Claude 3.7 and scored by a GPT‑4… | ||
| Forecast of Hourly Train Counts on Rail Routes Affected by Construction Work | Sebastian Folz, Maren Westermann | 31m |
DB Systel GmbH builds a model pipeline on AWS that converts legacy CSV to Parquet, engineers features with Pandas/Polars, trains LightGBM and PyTorch-Geometric… | ||
| Vector Streaming: The Memory Efficient Indexing for Vector Databases | Sonam Pankaj, Akshay Ballal | 25m |
Rust-based Embed-Anything implements vector streaming by moving embedding work into a separate thread/process and communicating via MPSC channels, keeping only… | ||
| What we talk about when we talk about AI skills. | Paula Gonzalez Avalos | 30m |
A data-driven claim: a 12-dimension AI skills framework and live visualization built from 269 survey responses at PyCon DE & PyData maps real competencies… | ||
| PosePIE: Replace Your Keyboard and Mouse With AI-Driven Gesture Control | Daniel Stolpmann | 21m |
PosePIE is an open‑source Python system that turns any webcam into a gesture input device using Ultralytics YOLO pose estimation with 17 COCO keypoints… | ||
| Optimizing Energy Tariffing System with Formal Concept Analysis and Dash | Irina Smirnova-Pinchukova | 30m |
Implements Formal Concept Analysis with FCAPy and an interactive Dash front end, producing linked object/attribute tables and a Hasse-diagram that enforces… | ||
| Towards Intelligent Monitoring: Detecting Degraded Flame Torch Nozzles | Dominik Falkner | 28m |
Uses structure-borne sound sensors mounted at the zone‑1 edge to avoid clipping, samples at 2 MHz sampling, and manual annotation yields 879 labeled patches… | ||
| Where have all the post offices gone? Discovering neighborhood facilities with Python and OSM | Katie Richardson | 29m |
Uses Python with PyOsmium, GeoPandas, Bokeh and Overpass/Geofabrik extracts to parse OSM .pbf snapshots, spatially join post office nodes to admin polygons… | ||
| From Rules to Reality: Python's Role in Shaping Roundnet | Larissa Haas | 32m |
Data pipeline uses Google Sheets + Colab to ingest and analyze annotated match video data, producing a baseline from 73 games annotated (17+ hours) that shows… | ||
| The Forecast Whisperer: Secrets of Model Tuning Revealed | Illia Babounikau | 24m |
Compares MAE, MSE and WMAPE on sparse retail series and shows MAE training produces a -6,000 profit while switching loss to MSE yields +3,000 and tuning for… | ||
| Responsible AI with fmeval - an open source library to evaluate LLMs | Mia Chang | 30m |
Open-source fmeval provides a structured pipeline of data loaders, model runners, evaluation algorithms and reports to move LLM work from experiments to… | ||
| The Future of AI: Building the Most Impactful Technology Together | Leandro von Werra | 54m |
Argues that fully open models are achievable and necessary, showing concrete work: rebuilding DeepSeek’s R1 training pipeline as Open R1, publishing FineWeb… | ||
| Electify - Retrieval-Augmented Generation for Voter Information in the 2024 European Election | Christian Liedl | 30m |
Electify is an open source RAG pipeline that converts 70+ page German party manifestos into question‑answer pairs by PDF parsing with pypdf, 1,000‑character… | ||
| Code & Community: The Synergy of Community Building and Task Automation | Cosima Meyer | 26m |
A working automated megaphone that posts PyLadies content to BlueSky and Mastodon, draws from an RSS-maintained registry, and uses GitHub Actions to run a cron… | ||
| What do a tree and the human brain have in common-a not so serious introduction to digital pathology | Daniel Hieber | 29m |
Whole-slide pathology images reach 100,000 x 100,000 pixels with three color channels (≈5.5 billion inputs), so the practical wins come from preprocessing not… | ||
| nothing matches | ||