PyData Seattle 2025
2025-11-07..2025-11-09
Bellevue, United States
[PyData Seattle 2017, PyData Seattle 2023, PyData Seattle 2025]
| title | author | length |
|---|---|---|
| Communicating Concepts, Clearly Explained!!! | Josh Starmer | 50m |
Explains three concrete goals and shows how they guide a single PCA Step-by-Step video that uses simple gene/mouse examples, hand-drawn graphs… | ||
| Real-TIme Context Engineering for Agents | Jim Dowling | 40m |
Claims a fast alternative to RAG: use application IDs (user/session/article/order) to fetch targeted context within seconds rather than retrieve text chunks… | ||
| Building valuable Deterministic products in a Probabilistic world | John Carney | 38m |
Argument: deliver valuable deterministic products while using AI only where it reduces cost or enables capability that deterministic code cannot. Concrete… | ||
| Optimizing AI/ML Workloads | Saurabh Garg | 40m |
Claims a concrete, deployed system: Metaflow for transparent usage reporting and cost attribution that instruments Kubernetes/Argo runs, captures node labels… | ||
| Driving Data Democratization | Zaheera Valani | 42m |
Claims that governance, sharing, and intelligent analytics actually scale: Unity catalog provides a single governance layer across tables notebooks dashboards… | ||
| Polars on Spark | Allison Wang, Shujing Yang | 31m |
Claims that combining PySpark Arrow UDFs with Polars yields real speed and memory wins: Polars is a Rust DataFrame that reads Arrow without copying, enabling… | ||
| Multi-Series Forecasting at Scale with StatsForecast | Khuyen Tran, Yibei Hu | 40m |
Retail demand pipelines use StatsForecast for fast statistical forecasts with rolling-origin cross-validation and conformal prediction intervals; model… | ||
| We don't dataframe shame: A love letter to dataframes | Devin Petersohn | 41m |
Claims a formal data frame model and an algebra that reduces pandas’ ~600 functions to 15 operators, shows how parallelism splits into row-wise and column-wise… | ||
| Data Loading for Data Engineers | Weston Pace | 34m |
Describes concrete changes: treat ordering as a first-class artifact with an explicit permutation table (row ids + split ids) instead of implicit per-worker… | ||
| Explore Solvable and Unsolvable Equations with SymPy | Carl Kadie | 34m |
Python with SymPy tests families of equations and reports what closes and what only numerics find. Polynomials degree ≤4 produce closed forms but degree ≥5… | ||
| Red Teaming AI: Getting Started with PyRIT | Roman Lutz | 44m |
PyRIT claims to automate and semi‑automate AI red teaming by generating attacks with an adversarial fine‑tuned LLM and scoring outputs with LM‑based judges… | ||
| Know Your Data(Frame) with Paguro | Bernardo Dionisi | 39m |
Paguro claims to be a lightweight, Polars-native validation and metadata library that uses Polars-native predicates (expressions) as checks, supports single… | ||
| Generalized Additive Models: Explainability Strikes Back | Pedro Albuquerque | 41m |
Claims concrete wins: GAM variants match or beat black‑box models on tabular data, show up in Kaggle winners and a study of 68,000 models, and a case where a… | ||
| Actually using GPs in practice with PyMC | Bill Engels | 44m |
PyMC's GP tools are shown as practical: a hierarchical model combines team, player and season effects with an age GP and yields a clear age–SLG peak at 26. The… | ||
| Wrangling Internet-scale Image Datasets | Kieran Merchant, Suarez | 33m |
Released dataset Re-LAION-Caption19M is a concrete claim: 19-million images with improved captions created by filtering multibillion-scale metadata… | ||
| Building Data-Driven Startups with User-Centric Design | Yinhan Liu, Pedro Luraschi, Joshua Ahmed, Yujian Tang | 40m |
Practical startups and concrete outcomes dominate: a platform that abstracts AWS to help nontechnical users launch AI products, an AI-driven social marketing… | ||
| Never Send a Human to do an Agent's Search | Chang She | 45m |
Claim: agents succeed or fail on data, not models. Demonstrated build: an open source Lance file/table format that stores multimodal assets, embeddings, and… | ||
| How to Optimize your Python Program for Slowness | Carl Kadie | 36m |
Short Python programs emulate Turing machines and also a stack of nested, dynamically bound loops to reach astronomic runtimes, including a Busy Beaver 5… | ||
| Taming the Data Tsunami | David Aronchick | 37m |
Claims a concrete open‑source playbook that shifts filtering and governance upstream and policy-as-code for distributed pipelines that reduce dataset size by… | ||
| Supercharging Multimodal Feature Engineering | Jack Ye | 42m |
Claim: Lance plus Ray actually lets pipelines add new features by writing new column files not by rewriting whole tables and run UDFs across fragments with Ray… | ||
| Lightning Talks | 38m | |
A rapid sequence of concrete outcomes and tools: a Haskell dataframe called dataf frame using types to catch nulls at compile time, a Python… | ||
| The Missing 78% | Noor Aftab | 40m |
Claims: women hold 22% of data and AI roles and make 3% of Python commits, a “missing 78%.” Methods: 1.5 years of mixed research, analysis of 24 groups, and a… | ||
| Prompt Variation as a Diagnostic Tool | Aziza Mirsaidova | 32m |
Claim: small prompt changes collapse accuracy — examples show 95% to 75% drop on rephrasings and Microsoft finds a 40% shift for GPT‑3.5 Turbo between markdown… | ||
| Unlocking Parallel PyTorch Inference (and More!) | Trent Nelson | 44m |
Claim: PyTorch inference runs concurrently inside the same process using Python free-threading. Built a multi-threaded pure-Python Paralleledia web stack and a… | ||
| Practical Quantization in Keras | Jyotinder Singh | 48m |
Quantization in Keras converts FP32 weights to lower-bit formats using Keras’s unified quantise API and KerasHub models, yielding measurable tradeoffs: weight… | ||
| Explainable AI for Biomedical Image Processing | Ojas Ankurbhai Ramwala | 46m |
Conference claim: over thousand FDA approved AI algorithms exist but clinical uptake lags. Built a validation platform called Glenwell AI to run external model… | ||
| Scaling Background Noise Filtration for AI Voice Agents | Stephen Cheng | 35m |
Claims a production rollout of background-noise filtration for hospital call agents that measurably reduces transcription errors and UX failures. Uses an… | ||
| Building Inference Workflows with Tile Languages | Andy Terrel | 31m |
Claim: inference gets 15x lower cost per token on recent Blackwell hardware and a tile compiler maps NumPy-style array code to tensor cores. Tools actually… | ||
| Democratizing (Py)Data | C.A.M. Gerlach | 32m |
A new open source remote development architecture for Spyder 6.1 actually lets a local IDE open consoles, variable explorer, plots and profiler on a distant… | ||
| Beyond Just Prediction: Causal Thinking in Machine Learning | Avik Basu | 43m |
If prediction and randomized AB tests disagree, causal methods estimate the effect of actions not just outcomes. Uplift modeling uses observed numbers like 20%… | ||
| There and back again... by ferry or I-5? | JustinCastilla | 41m |
A concrete build that uses the Model Context Protocol (MCP) to fetch WA Dept of Transportation ferry schedules, Google Maps drive times and live traffic, and… | ||
| Building Agents with Agent Bricks and MCP | Denny Lee | 40m |
Agent Bricks claims to let teams build production agents quickly by wiring a vector index of protocols plus PDFs and regulatory docs into a no‑code knowledge… | ||
| Why Models Break Your Pipelines | Everett Kleven | 36m |
Claim: models break pipelines because they’re expensive, stateful, and hard to configure. Concrete evidence: Daft adds stateful UDFs (async, generators… | ||
| Evaluation is all you need | Sebastian Duerr | 43m |
Maps the open‑source evaluation landscape, compares techniques like RAGAS and Evaluation Driven Development and frameworks such as DeepEval and Phoenix and… | ||
| Securing Retrieval-Augmented Generation | Rajesh Kumar | 33m |
Researchers and OWASP flag vector databases as a new attack surface and give concrete findings: embedding inversion can recover 50–80% of tokens from… | ||
| nothing matches | ||