PyData DC 2018
who it's for
data scientists, python developers, data engineers, machine learning engineers, analytics engineers, data visualization engineers
topics
pandas data cleaning and reshaping; Dask distributed dataframes; Airflow orchestration pipelines; TensorFlow model serving; Skorch scikit-learn + PyTorch; NLP transfer learning with GPUs; RAPIDS GPU data processing; interactive dashboards with Dash; plotly and Bokeh visualizations; web scraping with Beautiful Soup; probabilistic programming applications; algorithmic fairness measurement
| title | author | length |
|---|---|---|
| Keynote: The Business of Open Source | Travis E. Oliphant | 01h 01m |
Argues that open source succeeds or fails on concrete economics not slogans, citing SciPy hits 1.0 after 17 years and NumPy ~700 contributors as evidence of… | ||
| Keynote: A Sailor's View of Python And the idea of "Defensive Programming" | Steven Lott | 44m |
Argues that reliable Python systems need explicit design envelopes, detection, and practiced recovery not vague best practices, citing a midnight diesel leak… | ||
| Keynote: Using Data Science to Recruit Data Scientists | Maryam Jahanshahi | 52m |
Uses NLP and behavioral experiments on 15 million job descriptions plus candidate search logs to show concrete effects: adding senior data scientist cuts… | ||
| Keynote: Learning from Machines | Ashi Krishnan | 42m |
Shows concrete visualizations from Inception deep dream gradient ascent and explains convolutional filters, receptive fields, and how gradient ascent produces… | ||
| Recreating, Understanding, and Visualizing FiveThirtyEight's... | Matthew Brems, Joseph Nelson | 01h 42m |
Recreates FiveThirtyEight's elections forecast step by step and names the actual pieces: polls + fundamentals + experts + noise, implements pollster ratings… | ||
| Building a Machine Learning Enabled Bot in the Cloud | Jerry Hargrove | 01h 21m |
A working serverless Slack bot that classifies aircraft images using AWS Lambda, API Gateway, S3 and a SageMaker endpoint plus optional Amazon Rekognition is… | ||
| Scaling PyData with Dask | Martin Durant, Jim Crist | 01h 17m |
Hands-on tutorial shows how to scale Pandas and scikit-learn using Dask on a cloud Jupyter/Kubernetes cluster, actually launching a cluster with 20 workers, 2… | ||
| Cleaning and Tidying Data in Pandas | Daniel Chen | 01h 27m |
Shows concrete cleaning steps using pandas: loading files with read_csv, inspecting df.head/df.info, reshaping with pd.melt() to turn wide tables into long… | ||
| Five Kinds of Python Functions | Steven Lott | 01h 31m |
Defines five concrete function forms and shows what each actually buys: plain defs for single-expression returns, higher-order map chains are lazy with… | ||
| Web Scraping with Beautiful Soup | Monica Puerto | 58m |
Extracts IMDB pages using Beautiful Soup with requests + lxml and builds a reusable function that takes an IMDB ID and returns a dictionary of metadata… | ||
| Going beyond “Sorry, I didn’t get that” building AI assistants that scale... | Justina Petraitytė | 01h 31m |
Workshop shows how to build a machine‑learning assistant with Rasa NLU and Rasa Core, using Jupyter/Colab, CRF for NER, an SVM-style intent classifier and a… | ||
| Sell cron, buy Airflow: Modern data pipelines in finance | James Meickle | 39m |
Quantopian replaces brittle crontabs with Apache Airflow pipelines in less than six months, turning scattered scheduling logic into DAGs, checkpointed S3… | ||
| Inside Out - Creating a Python Ecosystem for Analysis and Better Government | Babila Lima | 42m |
Baltimore City builds a Python ecosystem that automates internal service workflows, produces reproducible reports, and changes hiring and training: a… | ||
| Right Code, Right Place, Right Time | Tim Hopper | 39m |
Argues that most data science problems are actually software engineering problems and that success turns on concrete choices: dependency management… | ||
| Intake - taking the pain out of data access | Martin Durant | 34m |
Intake provides a concrete claim: a single API that turns catalog entries into usable data with a one-liner .read returns a pandas DataFrame, backed by YAML… | ||
| Tracking plasma flows at the solar surface with the Solar Dynamics Observatory | Raphael Attie | 39m |
SDO delivers 4k x 4k images at ~1.5 TB per day and Python code implements a vectorized ball tracking optical‑flow method that settles thousands of simulated… | ||
| Webscraping Responsibly for Social Science Research | Graham MacDonald | 40m |
Urban Institute data team describes concrete methods and outcomes: Python requests and Selenium with an add-on called site monitor throttle requests and… | ||
| Avoiding Bad Database Surprises: Simulation and Scalability | Steven Lott | 35m |
Argues that prevention is possible by simulate early and simulate often, using Python to build a tiny processing nucleus, a JSON Schema–driven generator of… | ||
| The First Notebook War | Martin Skarzynski | 25m |
A public spat claims notebooks break basic software engineering—modularity and code reuse—while proponents show concrete fixes: JupyterLab as an IDE, Jupytext… | ||
| Skein: a simpler way to deploy applications on Hadoop Clusters | Jim Crist | 26m |
Skein is a small library and CLI that actually runs Python apps on Apache YARN by shipping a Java application master that speaks gRPC to Python clients and… | ||
| uarray - Efficient and Generic Array Computation | Travis E. Oliphant, Saul Shanabrook | 43m |
Claims a practical system uarray that rewrites array expressions via the Mathematics of Arrays to eliminate temporaries and unneeded work, then emits code for… | ||
| Measures and Mismeasures of Algorithmic Fairness | Manojit Nandi | 41m |
Concrete failures and fixes instead of abstraction are the deciding factors here: Google Cloud Vision mislabels gender and Google Photos once tags Black people… | ||
| Task Failed Successfully | Jeremiah Lowin | 36m |
Talk claims modern pipelines break when workflow systems treat failure as an outlier and documents concrete Airflow pitfalls: XComs are a cheat (admitted… | ||
| Skorch - A Union of Scikit-learn and PyTorch | Thomas J Fan | 38m |
A scikit‑learn–compatible wrapper called skorch actually hides the training loop and exposes fit / partial_fit / set_params so neural nets plug into pipelines… | ||
| Machine Learning in the age of increasing Data Privacy Consciousness | Sam Talasila | 33m |
Argues concrete costs and practices: cites 20 million or 4% of global revenue as GDPR fines and 87% identifiable from date of birth gender and zip as a risk… | ||
| Single Page Apps: What an API Developer Should Know | Alison Rowland | 29m |
Eight months leading a front-end team building an Angular client uncovers concrete tradeoffs: deploy SPA bundles statically (S3) while the API must enforce… | ||
| Mining dockless bikeshare and dockless scootershare trip data | Stefanie Brodie, Kiana Roshan Zamir | 45m |
City permit data plus company monthly reports and public APIs feed Python scripts that produce concrete answers not abstracts: a random forest classifier… | ||
| Democratizing NLP content modeling with transfer learning using GPUs | Sanghamitra Deb | 43m |
Claims building in-house embeddings with using GPUs to enable transfer learning across tasks like recommendation, customer insights, and content-coverage… | ||
| RESTful Machine Learning with Flask and TensorFlow Serving | Carlo Mazzaferro | 41m |
A Python library that takes Keras models from notebook to production by scaffolding Docker and TensorFlow Serving, exposing a Flask + Swagger REST API that… | ||
| Do Your Homework! Writing tests for Data Science and Stochastic Code | David Waterman | 42m |
Concrete method for testing nondeterministic code: write gold tests (human-labeled JSON examples), use Pytest examples with fixtures and xfail, and benchmark… | ||
| Probabilistic Programming in the Real World | Zach Anglin | 43m |
Shows concrete models and results: uses PyMC3 and MCMC variants (Metropolis, HMC/NUTS) to build a truncated-normal joke-rating model from three ratings, a… | ||
| Data visualization with plotly py: Version 3 and beyond | Jon Mease | 37m |
Plotly.py version 3 adds two-way widget sync so figures become live FigureWidget objects with an imperative property-assignment API, real-time hover and… | ||
| Approaching AI in Banking | Hussain Sultan | 37m |
Argues that AI in banking centers on operationalizing unit economics and segmentation into an auditable decision pipeline rather than flashy chatbots, with a… | ||
| Words in Space: A Visual Exploration of Distance, Documents, and Distributions... | Rebecca Bilbro | 45m |
Argues that picking a distance metric matters more than the topic name and shows concrete comparisons: Euclidean, Manhattan, Chebyshev, Minkowski, Mahalanobis… | ||
| Making interactive visualizations easy, inside and outside of notebooks | Julia Signell | 36m |
Shows a concrete stack and what it actually does: uses hvPlot, HoloViews and GeoViews with datashader and xarray to render millions of points in the browser… | ||
| Accelerating Data Science with RAPIDS | Keith Kraus | 39m |
Claims GPU-accelerated end-to-end pipelines using cuDF and cuML with Apache Arrow memory sharing, Dask for distribution, and CUDA-backed JIT for Python UDFs… | ||
| Dimensionally reducing data squeezing out the good stuff | Aabir Abubaker Kar | 34m |
Explains which algorithms actually reduce dimensions: k‑means as distance clustering, PCA via variance (example directions explaining ~30 vs ~1.4 variance) and… | ||
| Developing Dashboard Applications using Bokeh | Luke Canavan | 26m |
Describes building a real dashboard: reproduces lessons from the Dask Distributed diagnostic UI, constructs a live Bokeh server app that performs OpenCV face… | ||
| End to End Machine learning pipelines for Python driven organizations | Nick Harvey | 39m |
Claim: pachyderm gives data provenance by assigning commit IDs to every dataset change so every model input and output is traceable. Tools actually used are… | ||
| Using Sockeye Neural Machine Translation in a Streaming Pipeline | Jeff Zemerick | 35m |
A concrete pipeline that consumes live tweets, detects language with Apache OpenNLP, sends sentences via Apache Flink and Thrift to a Sokka neural model and… | ||
| Dash: data exploration web apps in pure Python | Chelsea Douglas | 41m |
Claim: build interactive data-exploration web apps in pure Python using Dash, built on plotly.js, Flask and React, with callbacks that allow multiple inputs… | ||
| Why do you have to make it so complicated? | James Powell, Brandon Burroughs, Daniel Allan | 49m |
A practical framework and two‑page handout maps concrete tradeoffs: when code moves from notebook→module→package to multi‑repo; when to adopt git, GitHub, git… | ||
| Graphs: Datastructures to Query | Benjamin Ortiz Ulloa | 43m |
Survey of concrete recipes and results: uses igraph Python/R/C/C++ and an iPython notebook on GitHub, shows a graph-based recommendation built from bipartite… | ||
| Beautiful, Interactive, and Portable Maps using Folium and Live API Data | Ariel M’ndange-Pfupfu | 38m |
Shows how to combine Folium + Leaflet.js with live APIs (Census, WMATA, Walkscore) plus Flask and simple caching so Python produces HTML/JS maps that render in… | ||
| nothing matches | ||