PyData Vermont 2025
as the organisers put it
PyData Vermont 2025 is a two-day in-person conference for data scientists, engineers, and developers, featuring keynotes, talks, and networking at UVM in Burlington, VT, October 21-22.
who it's for
data scientists, data engineers, python developers, machine learning engineers, natural language processing practitioners, geospatial analysts, data visualization designers, ml infrastructure engineers
topics
open source document ingestion; OCR with visual grounding; hybrid chunking for long docs; multimodal retrieval augmented generation; vector search with Milvus; address matching and record linkage; GPU-accelerated PyData with RAPIDS; cuDF and cuML workflows; reproducible Python environments with Conda; MCP servers and Claude integrations; tokenization and distributional diagnostics; GeoPandas spatial joins and reprojection
| title | author | length |
|---|---|---|
| Complex Data Ingestion with Open Source AI | Zhao | 01h 00m |
Claims an open-source toolkit named Docling reliably converts messy PDFs, scans and slides into structured JSON/markdown/HTML, preserving headers, tables… | ||
| Cleaning Messy Data at Scale: APIs, LLMs, and Custom NLP Pipelines | Dody | 48m |
Example uses 5 million raw US addresses matched to a 120 million ground truth; simple left join starts at 47% match rate, adding regex + open-source parser… | ||
| GPU-Accelerated Data Science for PyData Users | Clementi, McCarty | 16m |
RAPIDS shows how to move real pandas and scikit‑learn workflows onto NVIDIA GPUs with cuDF and cuML, using the same PyData APIs so code changes stay minimal… | ||
| MCP basics with Conda and Claude | Bouquin | 56m |
Builds a working MCP server in Python and a reproducible conda env that lets Claude Desktop call the New York Times Books API for real‑time bestseller data… | ||
| Context is all you need: FUNdamental linguistics for NLP | Zimmerman, Ashley | 46m |
Argues that distributional hypothesis explains much of LLM competence, shows concrete tools—alotaxonometer, shiftator, labMT lexicon and Python packages—that… | ||
| From Chaos to Confidence: Solving Python's Environment Reprodu... | Wages | 30m |
Clear decision hinges on specifics: what exact reproducibility claim is made, which tools are used (Pixie, pip, conda, mamba, poetry), and whether the workflow… | ||
| A Practical Introduction to Geospatial Python | Smith | 01m |
Claims an executable end-to-end geospatial pipeline that goes from acquiring open data through cleaning and spatial joins to making maps and simple dashboards… | ||
| Open Source Vermont Data Platform: Access, Analysis, and Visualization | Koch | 41m |
Prototype for the Vermont Data Collaborative at UVM Center for Rural Studies centers on concrete decisions: cleaner pipelines in Python, a Streamlit prototype… | ||
| The Art of Data: Hand-crafted, Human-centered Dat... | Fortney, Cooley | 19m |
Hands-on workshop on small data that builds physical artifacts from five preselected datasets using markers, stickers, yarn and simple visual encodings to… | ||
| nothing matches | ||