Data By the Bay 2016
as the organisers put it
Spanning a hundred of talks over 3 days on May 2017, Data By the Bay is a by-data engineers, for-data engineers developer and data scientist conference.
| title | author | length |
|---|---|---|
| Apache Beam (incubating) | Eugene Kirpichov | 21m |
Apache Beam traces Google’s evolution from MapReduce/FlumeJava/MillWheel into a single programming model that unifies batch and streaming, implemented as Java… | ||
| Predicting Hacker News with Beam and TensorFlow | Eric Anderson | 21m |
Demonstrates using Apache Beam and Google Cloud Dataflow with Java/Python SDKs to solve stragglers and cost by dynamic work rebalancing (aka liquid sharding)… | ||
| Recommendations for Building Machine Learning Software | Justin Basilico | 41m |
Netflix runs recommendations for 80 million members in 190+ countries and reports 80% of what people watch comes from recommendations. The system uses AWS… | ||
| Deep dive and best practices of Spark streaming | Prakash Chockalingam | 40m |
Explains Spark Streaming internals and practical fixes: receivers buffer data on executors, direct Kafka avoids receivers so scale by increasing Kafka… | ||
| Hidden GEMMs: How Optimized Math Libraries Work | Marek Kolodziej | 40m |
Shows how a naive C triple loop yields ~1,200 ms while simple loop reordering cuts that to ~167 ms and cache-aware tiling plus ICC brings it to ~56 ms (NumPy… | ||
| Data in the Apache "big data" ecosystem | Konstantin Boudnik | 20m |
Argues that most enterprises do not need massive clusters: a Forrester survey of ~1,800 firms shows less than 10% at petabyte scale and typical corporate… | ||
| data.bythebay.io: Shubho Sengupta, | Shubho Sengupta | 42m |
Claims a practical end-to-end deep speech system that replaces complex ASR pipelines, trained on 12,000+ hours (synthetic to ~100,000) and delivering… | ||
| A Match Made at Upwork | Alya Abbot | 20m |
Upwork runs a content‑based Job Interest model (simple logistic regression) that predicts whether a freelancer will apply to a job using profile, past… | ||
| Automating Data Science for the IOT | Ruban Phukan | 20m |
Describes an automated pipeline that turns noisy, irregular IoT sensor streams into predictive maintenance models by programmatic feature engineering… | ||
| Detecting Anomalies in Streaming Data | Subutai Ahmad | 43m |
An open-source HTM implementation processes each timestamped value in ~10 ms, continuously learns temporal sequences and emits a normalized anomaly score plus… | ||
| The Security Wolf of Wall Street | Thibault Reuille | 45m |
OpenDNS describes a live pipeline that turns ~80B DNS queries/day into real-time blocks by running a Python + ZeroMQ Avalanche stream framework across EC2… | ||
| Finding Your Audience In The Internet of Things | Dhruv Choudhary | 20m |
Automatic Labs collects OBD data and compresses trips into velocity-acceleration frequency histograms, denoises them with a stacked denoising autoencoder, then… | ||
| Data and Society | Alexy Khrabrov, Brad Newman, Elizaveta Malashenko, Gurinder Sangha, Nicole Shanahan | 01h 06m |
Panel documents concrete wins and hard limits: a startup Lit IQ trains computational linguistics on annotated contracts and case law to surface drafting bugs… | ||
| The Promise of Heterogeneous Computing | Subbu Rama | 18m |
Software grows more abstract while hardware fragments into GPUs, TPUs, FPGAs, DSPs and memory‑adjacent chips, creating a gap; Google ships a TPU for TensorFlow… | ||
| Building an A.I. Cloudy Sky: What We Learned at PredictionIO | Simon Chan | 38m |
PredictionIO is an open-source ML server (Scala/Java) that surfaces ten concrete platform decisions: event server for mixed batch/stream imports, three real… | ||
| Deep Dive: Spark Memory Management | Andrew Or | 41m |
Covers how Spark actually divides JVM heap between execution and storage and the change to unified memory management (Spark 1.6) that lets execution borrow… | ||
| Building AI that Searches and Sells | Anand Kulkarni | 21m |
LeadGenius builds a data pipeline that scrapes ~55 sources using Scrapy and BeautifulSoup plus browser automation (Selenium), running on an Azure cluster that… | ||
| Building a Graph of all US businesses using Spark technologies | Alexis Roos | 21m |
Builds a production business index using Spark GraphX with Scala and MLlib models and claims it resolves noisy, duplicate records into real companies by… | ||
| Making Predictions Under Lending Regulations | Yu Jay Huoh | 21m |
Earnest trains separate classifiers for each decline reason using credit reports plus transaction history, builds each as a random forest (≈500 trees), and… | ||
| Case Law and ML on Spark | Jeremy Corbett | 21m |
Migration from Apache Pig to Spark powers a pipeline that processes a legal corpus of about 7 million opinions and 54 million citations, running NER, LDA… | ||
| Hidden in plain sight: Using law to summarize the law | Richard Downe | 23m |
Case law gets automatically summarized by harvesting judge-written parenthetical summaries and stitching frequently quoted text into key passages, using… | ||
| End-2-End Monitoring and Troubleshooting a Real-Time Data Pipeline | Alan Ngai | 22m |
A monitoring platform that centralizes Kafka, Spark, Storm, Flink, Cassandra and Elasticsearch metrics claims to make pipeline debugging measurable by modeling… | ||
| Data From Click To Settlement: A Full-Funnel UX Approach | Michael Terry | 22m |
Lawfty builds a pipeline that links ad click to case settlement and uses concrete tactics not buzzwords. It serves unique phone numbers via Twilio API per… | ||
| Synthesizing human and machine capabilities | Eric Colson | 18m |
Claim: stitching machines and humans yields better recommendations than either alone. The system runs PCA and SVD plus matrix factorization and neural nets on… | ||
| Data governance and distribution with Dat | Karissa McKelvey | 30m |
A three-person team (expanding to seven) builds Dat: a Node.js, peer-to-peer system using hyperdrive, discovery-swarm and Rabin fingerprinting to chunk files… | ||
| Analytics as Code with Juttle | Daria Mehra | 23m |
Argues that operational analytics must be treated as code to get versioning, reviews and reuse; shows concrete problems: Kibana-backed setups balloon to over a… | ||
| Quantifying Democracy and Freedom with Human Rights Metrics | Dimitri Gusev | 36m |
Compares country-level fertility (World Bank) to girls’ upper secondary completion (UNESCO, Barro‑Lee) and to Democracy Index and Freedom House scores using… | ||
| Gathering around the data table | Jason Lally | 16m |
DataSF centers use not just publication: the city publishes datasets at data.sfgov.org API (JSON, CSV, GeoJSON) and exposes table-level API docs; teams built a… | ||
| Breaking Down Paywalls for Online Health | Pauline Ng | 33m |
About 26% of people searching health info hit paywalls, articles cost about $32 each, and publishers have raised prices over 250%. A Java standalone called… | ||
| Privacy Issues in Big Data Processing in light of Data Breaches | S. Srinivasan | 26m |
Five high-profile breaches show the attack vectors and failed protections: third-party credentials let attackers pivot into Target and Home Depot, FireEye… | ||
| Single Customer View | Thomas Trolez | 17m |
A Single Customer View is built end-to-end here: a Play server captures events, pushes them to Kafka → S3 + real-time processors, Spark builds sessionization… | ||
| Text Analytics Simplified | Trung Diep, Ronald Sujithan | 24m |
Data Ninja’s Smart Content API returns JSON with extracted concepts, categories, entities, keywords and sentiment driven by a continuously updated knowledge… | ||
| Black Magic: How to apply Machine Learning to real-world problems | Evion Kim | 46m |
Mattermark builds an automated funding-extraction pipeline that turns news articles into structured records (amount, series, investors, date) and pulls 2015… | ||
| Building semantic search using Deep Learning | Samiur Rahman | 32m |
Semantic search replaces token matching with vector similarity: a REST ML service holds precomputed vectors for ~1.3 million companies (trained on ~4 million… | ||
| Unsupervised NLP Classification with Clustering | Matthew Seal | 39m |
Pipeline processes budget PDFs with NLTK tokenization, stemming and stopword filtering, builds minhash approximation indexes, computes Jaccard edges above a… | ||
| Why The Best Minds Of Our Generation Are Thinking About Click Ads | Josh Wills | 32m |
Argues that ad tech created the best data tooling and culture but is now mostly an automated machine war; identifies the real data problem as change not scale… | ||
| Open Collaboration for Civic Impact | Lawrence Grodeska | 22m |
CivicMakers argues that civic tech needs a reproducible lifecycle—Discovery, Alpha, Beta, Live—using co-discovery street interviews, low‑fidelity prototypes… | ||
| Design for Local Government: Data, Services, and Transparency | Steve Pepple | 20m |
OpenGov builds a SaaS platform that ingests messy general ledgers, reverse-engineers chart of accounts hierarchies, and produces searchable visualizations… | ||
| Hot Legal Issues | Francoise Gilbert | 39m |
Overview of legal constraints on big data with concrete examples: the Fair Credit Reporting Act can make analytics into regulated credit reports, the Spokeo… | ||
| An approach to internal search query analysis | Max Ho | 20m |
Course Hero analyzes >9M content items by parsing raw search strings into keywords and preset study-guide pairs, grouping 1D/2D/3D terms and plotting a demand… | ||
| Human Intent and Text Understanding | Alexy Khrabrov, Jenny Finkel, Robert Munro, Surabhi Gupta, Tihomir Bajić | 01h 04m |
This talk video is a node in devreal.ai, the developer community graph.
It connects to:
Talk: Panel: Human Intent and Text Understandi... —… | ||
| Data and Algorithmic Bias in the Web | Ricardo Baeza-Yates | 40m |
This talk video is a node in devreal.ai, the developer community graph.
It connects to:
Talk: Data and Algorithmic Bias in the Web —… | ||
| lda2vec | Chris Moody | 49m |
Describes a concrete lda2vec hybrid that adds a learned document vector to skip‑gram word2vec, converts document factors to proportions with softmax, and… | ||
| Mining Noisy Transaction Data with Neural Nets | Frank Taylor | 41m |
Company transaction logs total roughly 1.5 TB and 100M rows; 22,000 rows get hand-labeled in a labeling party to create ground truth. Text tokens are… | ||
| Aspect Based Sentiment Analysis in 20+ Languages | Antonio Valderrabanos | 45m |
This talk video is a node in devreal.ai, the developer community graph.
It connects to:
Talk: Aspect Based Sentiment Analysis in 20+ L... —… | ||
| Smarter Search with Spark, Solr and Machine Learning | Grant Ingersoll | 40m |
Shows how Apache Spark and Solr are glued to build ranking‑centric search: live Spark‑Solr integration (open source on GitHub), a commercial Fusion layer, and… | ||
| Data & Metadata at the Internet Archive | Greg Lindahl | 43m |
The Internet Archive stores about 25 petabytes of content including 482 billion web captures and adds roughly 2 billion captures per week; storage uses custom… | ||
| Host Sponsor Welcome: IBM Analytics d2 | Joel S. Horwitz | 08m |
Claims that Apache Spark is becoming the analytics operating system, IBM invested in Spark last year and Watson now runs on Spark, and Bluemix offers… | ||
| byte2vec: a flexible embedding model constructed from bytes | Parsa Ghaffari | 25m |
byte2vec builds word embeddings from raw bytes rather than tokens, producing language‑independent vectors that handle out‑of‑vocabulary and morphologically… | ||
| Time series analytics for Big Data and IoT with Kx | Finton Quill | 15m |
Claims that kdb+ (columnar time-series DB) unifies streaming, real-time and historical data and cuts developer and TCO by using in‑memory and on‑disk… | ||
| PhrazIt : Tool for automatic text summarization | Niyati Parameswaran | 19m |
PhrazIt performs extraction summarization by ranking sentences with context-based indexing instead of just term frequency/position/length, building sentence… | ||
| SAMEntics : Tools for paraphrase detection | Niyati Parameswaran | 31m |
SAMEntics claims unsupervised paraphrase detection and generation that combines augmented structural metrics (Jaro, Jaro‑Winkler, Dice, cosine, Levenshtein… | ||
| Of Rules and Probabilities: Computational Linguistics Methods | Vita Markman | 26m |
Builds a title grammar from ~100k profile titles to auto-label short search queries, induce a skills dictionary, add grammar-derived features to a CRF and… | ||
| An Innovative Approach to Labeling Ground Truth in Speech | Jill Desmond | 20m |
VersaTag GUI replaces CLAN and feeds audio through a speech activity detector, presents one acoustic “utterance” at a time, and forces categorical dropdowns… | ||
| Automatic Links for a Web of Text | Scott Waterman | 41m |
Argues that hyperlinks become part of the text itself and demonstrates dynamic links that the browser plus search engines and web archives can assemble to add… | ||
| Hunting Criminals with Hybrid Analytics | David Talby | 24m |
A practical system detects insider trading by combining separate models for NLP, graph linkage and time-series routines into an ensemble that retrains in… | ||
| Scalably Internationalizing Millions of Latent Semantic Labels | Xiao Fan | 33m |
Built a system called CLAWS n‑gram mapping that converts job titles into labels by matching n‑grams and trained it on 300,000 English titles → 30,000 labels… | ||
| Using Spark MLlib for NLP | Michelle Casbon | 34m |
Built a standalone system called IdiML that turns text into feature vectors, trains logistic‑regression models with Spark MLlib vectors (not DataFrames), and… | ||
| Image Retrieval using Short Texts | Maarten Bosma | 22m |
ImageSuggest sits between user text and external image search APIs, extracting ranked keywords from Whisper posts and feeding cached third‑party queries; it… | ||
| Functional Programming for Machine Learning and Data Pipelines | Malcolm Greaves | 40m |
Argument: functional programming plus a strong type system reduces implementation bugs and makes ML code reusable. Concrete tools shown: Scala type classes and… | ||
| Panel: Future of Data Services | Alexy Khrabrov, Haoyuan Li, Joel S. Horwitz, Marius Eriksen, Neha Narkhede, Roman Shaposhnik, William Morgan | 59m |
This talk video is a node in devreal.ai, the developer community graph.
It connects to:
Talk: Panel: Future of Data Services —… | ||
| Parallel and distributed big joins in H2O | Matt Dowle | 41m |
Radix-sort based joins are parallelized and distributed inside H2O (Java backend, R/Python front ends) so large-key integer joins run in memory across a… | ||
| Concord: Simple & Flexible Stream Processing on Apache Mesos | Shinji Kim | 28m |
Concord is a Mesos-integrated stream processor with a core in C++, operator isolation via Mesos containers, per-operator RocksDB local state, and three routing… | ||
| Real-time, Streaming Advanced Analytics.... | Chris Fregly | 45m |
Live demos show a working pipeline that ingests clicks with NiFi→Kafka→Spark Streaming→Cassandra, trains ALS collaborative filtering (saved to Elasticsearch)… | ||
| Building domain specific databases with a distributed commit log | Yungjin Xu | 30m |
Smyte replaces Redis with Kafka + RocksDB, writing timestamped increments to Kafka and storing count plus Kafka offset in the same RocksDB transaction to get… | ||
| Characterizing the performance of Big Data processing platforms | Manish Singh | 38m |
Compares Hadoop MapReduce terasort, Hive TPC-DS and Databricks Spark benchmarks on AWS and reports that CPU and memory scale roughly linearly but network… | ||
| Taming JSON with SQL: From Raw to Results | Greg Rahn | 36m |
Compares six systems—Hive, Drill, Spark, BigQuery, Redshift, Snowflake—by running the same JSON dataset through each and producing a three-row five-column… | ||
| Building a Realtime Receiver with Spark Streaming | Sal Uryasev | 28m |
Builds three concrete Spark Streaming receivers: a simple foreign-exchange HTTP API (used to join GoFundMe donation currency data and to backfill via file… | ||
| The Engineer's Guide to Streaming | Ilya Ganelin | 41m |
Compares Apache Storm, Flink and Apex by scalability, performance, durability and availability and claims those, not throughput numbers, decide platform fit… | ||
| Netflix Keystone - Streaming Data Pipeline @Scale in the Cloud | Monal Daxini | 20m |
Netflix Keystone moves over 700 billion events per day (peaked at a trillion) with at‑least‑once semantics, using fronting and consumer Kafka clusters (4000+… | ||
| Investing in Data Analytics | Kanu Gulati | 22m |
Claims that the era of zettabytes (10 ZB in 2016 rising toward 40 ZB by 2020) creates investable opportunities in Industrial IoT are grounded in three… | ||
| Apache Flink: A Very Quick Guide | Vladimir Bacvanski | 38m |
Shows Apache Flink as a unified stream-and-batch engine that runs Scala DataStream API and Java, uses RocksDB state backend for large operator state outside… | ||
| Quantitative Trading with Machine Learning | Martin Froehler | 47m |
Company Quantiacs presents hands-on results using MATLAB and Python toolboxes to build a neural network auto-regressive model that predicts next-day futures… | ||
| A Real Time Analytics Framework | Hiral Patel | 17m |
Yahoo builds a reusable real‑time analytics framework that surfaces events in seconds (goal: <1 minute) by combining Storm topologies for in‑memory… | ||
| Akka Streams for Large Scale Data Processing | Zack Loebel-Begelman, Dustin Lyons | 01h 20m |
This talk video is a node in devreal.ai, the developer community graph.
It connects to:
Talk: Akka Streams for Large Scale Data Proces... —… | ||
| Google BigQuery | Michael Entin | 34m |
This talk video is a node in devreal.ai, the developer community graph.
It connects to:
Talk: Google BigQuery: a fully-managed data an... —… | ||
| Time-series Feature Engineering Done Right | Parag Namjoshi | 35m |
Workday builds retention-risk models using a temporal validation pipeline: train on events from 6–3 months back, validate on the last 3 months, and slide or… | ||
| Building a Real-time Streaming Platform Using Kafka Streams... | Jay Kreps | 35m |
Argues that all data can be modeled as table–stream duality and that Apache Kafka is a durable commit-log acting as a central streaming backbone. Describes… | ||
| DBTB INT Daria Mehra | Daria Mehra | 03m |
Argues that operational analytics must be implemented as code not GUIs, describing a move from one-off analyses to reproducible analytics-as-code pipelines… | ||
| DBTB INT Samiur Rahman r | Samiur Rahman | 03m |
Claim that a custom semantic search for millions of documents often outperforms using Elasticsearch out of the box, with examples of what was built at… | ||
| DBTB INT Matthew Seal r | Matthew Seal | 03m |
Argues that data now answers problems people previously couldn’t solve and that combining domains — democracy, NLP, design, science — produces practical… | ||
| DBTB INT Dmitri Gusev r | Dmitri Gusev | 03m |
Introduces a small implementation of a developer community graph (devreal.ai) that links talks, interviews, people and institutions and shows how that graph… | ||
| DBTB INT Matt Mahan r | Matt Mahan | 04m |
Argues that data and engineering can scale civic participation by turning messy public records into usable tools, using the voter file, election results and… | ||
| DBTB IN Eric Colson r | Eric Colson | 02m |
Stitch Fix uses algorithms plus human stylists to deliver results that pure machine learning cannot: the claim is that hybrid human-machine systems let the… | ||
| DBTB INT Akash Mukherjee | Akash Mukherjee | 04m |
Claim: data is domain-agnostic and touches every industry, so visualization choices must adapt as data volume and tech change. Recommends keeping an open mind… | ||
| DBTB INT Aaron Kalb r | Aaron Kalb | 02m |
Argues that observing past human interactions with data lets software guide better future behavior and reduce spurious analysis, and demonstrates a working… | ||
| DBTB INT Ian Lin r | Ian Lin | 05m |
Identifies concrete business problems on realtor.com by parsing event logs and product metrics, then applies data cleaning followed by modeling and imputation… | ||
| DBTB INT Kyle Hailey r | Kyle Hailey | 03m |
JavaScript libraries like Highcharts and D3 finally make the interactive visuals long wanted, but dashboards built by generalist devs often just throw up… | ||
| DBTB INT Gopal r | Gopal | 02m |
Compares healthcare to defense, finance and social media and claims those shared methods are now essential to clinical workflows, arguing data is the solution… | ||
| DBTB INT Gurinda Sangha r | Gurinda Sangha | 02m |
Claims that lit IQ is built around using data to improve legal decision making and that law as last frontier means a huge untapped market for analytics… | ||
| DBTB INT Alyssa Morrow r | Alyssa Morrow | 04m |
Argues that existing genome viewers cannot scale to three petabytes of data and demonstrates work on distributed visualization for genomics that pushes… | ||
| DBTB INT James Mulholland r | James Mulholland | 03m |
UX design at Platfora reframes decisions around measurable checkpoints, arguing that remixing the design process means inserting data at understanding users… | ||
| DBTB INT Vinay Prabhu r | Vinay Prabhu | 06m |
Claims that a smartphone’s onboard sensors combined with signal processing and machine learning create a working passive phone authentication pipeline that can… | ||
| DBTB INT Brian Carlsen r | Brian Carlsen | 04m |
Argues that the National Library of Medicine’s UMLS is a giant, messy medical thesaurus worth using, that public datasets paid for by taxpayers need the… | ||
| DBTB INT Mark Tuttle r | Mark Tuttle | 03m |
Computing resources become cheap so data becomes the biggest lever for solving messy unsolved problems, and open content is predicted to follow open software… | ||
| DBTB INT John Whaley r | John Whaley | 04m |
UnifyID demonstrates passive authentication by combining phone and IoT sensors to extract gait cadence from accelerometer streams and uniquely identify people… | ||
| DBTB INT Daniel Doerr r | Daniel Doerr | 03m |
Manager-level data work at Credit Karma argues big data is a wild west of on-the-job learning and bootcamps, recommends Kaggle competitions as the best free… | ||
| DBTB INT Nish Bhat r | Nish Bhat | 02m |
Founding engineer of Color Genomics argues genomics data generation now outpaces analysis capacity and calls for hiring and training cross-disciplinary data… | ||
| DBTB INT Steve Pepple r | Steve Pepple | 03m |
A product designer describes working with urban data from San Francisco Planning, DataSF and OpenOakland, running small playful experiments with standard… | ||
| DBTB INT Daniel Golden r | Daniel Golden | 02m |
Work uses convolutional neural networks to build automatic segmentation for radiology images, addressing the prior lack of labeled data and tooling. It… | ||
| DBTB INT Brian Studwell r | Brian Studwell | 02m |
Ravel Law product design perspective argues that data behaves like data as currency, reshaping systems and social behavior, and that analytics often augment… | ||
| DBTB INT Sai Moturu r | Sai Moturu | 03m |
Head of data science at Ginger.io explains how analytics are embedded into a real-world mental health care product that pairs algorithmic signals with a human… | ||
| DBTB INT Hernan Asorey r | Hernan Asorey | 03m |
Claims that data science at Salesforce is being reworked to deliver faster, sustainable impact by applying basic engineering principles so small teams can do… | ||
| DBTB INT Boriana Viljoen r | Boriana Viljoen | 03m |
Claims that compacted data can become actionable and drive behavior change, arguing environmental datasets like tidal currents deserve more attention for user… | ||
| DBTB INT Petros Giannikopoulos r | Petros Giannikopoulos | 06m |
A founder of Driver explains how genomics turns pathology from analog interpretation into discrete data points so treatment choices become rule based rather… | ||
| DBTB INT Sanghamitra Deb r | Sanghamitra Deb | 05m |
Describes an NLP pipeline that extracts side effects and target age groups from public clinical text, not by running new trials but by mining existing records… | ||
| DBTB INT Atul Butte r | Atul Butte | 04m |
UCSF builds an Institute that treats hospital records plus genomic reads as precision medicine inputs, arguing these are some of the most expensive but mostly… | ||
| DBTB INT Elad Ferber r | Elad Ferber | 05m |
Echo Labs codifies a pragmatic playbook for using healthcare data: insist that each data scientist combines programming speed, statistical rigor and… | ||
| DBTB INT Katherine Ahern r | Katherine Ahern | 04m |
Manager of visualization and analytics explains why deadlines forced concrete deliverables and why predictive analytics plus UX on large data still lacks… | ||
| DBTB INT Eric Williams r | Eric Williams | 05m |
Claims that the same analytics, machine learning and experimentation stack used to optimize clicks can be repurposed to optimize health outcomes, replacing ad… | ||
| DBTB INT Alan McLean r | Alan McLean | 02m |
Argues that health data visualization succeeds only when teams define the message first and map visuals to specific user goals, using Fitbit product work as… | ||
| DBTB INT Chester Chen r | Chester Chen | 05m |
Demonstrates how to instrument ML algorithms to stream real-time ML visualizations that reveal cost function shape and convergence behavior during long runs… | ||
| DBTB INT Hunter Whitney r | Hunter Whitney | 04m |
Argues that user experience and machine learning enter a tight, unpredictable coevolution of humans and machines where interfaces both observe and adapt, not… | ||
| DBTB INT Jeremy Howard r | Jeremy Howard | 04m |
Argues that deep learning already changes everyday products by powering Google Photos face recognition, Instagram-style automatic image edits and Google Inbox… | ||
| DBTB INT Andy Petrella | Andy Petrella | 07m |
A platform for interoperable, scalable genomics and health data that moves teams from ad hoc analysis to production-ready pipelines by pushing curated datasets… | ||
| DBTB INT Zachary Taschdjian r | Zachary Taschdjian | 03m |
Product manager for data visualization and interaction design at General Electric Digital describes the Industrial Internet and data combining into what GE… | ||
| DBTB INT Alya Abbot r | Alya Abbot | 06m |
Upwork has near-complete visibility of the hiring flow from job view to contract close and uses that complete hiring funnel to pick measurable objectives and… | ||
| DBTB INT Marek Kolodziej r | Marek Kolodziej | 05m |
Argues that current datasets are good enough to build real products but CPU cycles are wasted when people run linear algebra naively, so using efficient linear… | ||
| DBTB INT Ruban Phukan | Ruban Phukan | 03m |
DataRPM positions itself to automate data science for industrial use, focusing on machine data and IOT where the claim is opportunities reach trillions of… | ||
| DBTB INT Lukas Biewald r | Lukas Biewald | 02m |
Says data volume and access have grown so many projects now start by finding previously unused public sources and building simple pipelines to test ideas… | ||
| DBTB INT Subutai Ahmed r | Subutai Ahmed | 03m |
Argues that IoT streaming makes conventional approaches fail: traditional ML doesn't scale for continuously arriving sensor data, so detection and pattern… | ||
| DBTB INT Eugene Kirpichov r | Eugene Kirpichov | 04m |
A Google Cloud Dataflow team builds and ships a production feature called No Shard Left Behind, a dynamic worker balancing system for Dataflow that took about… | ||
| DBTB INT Scott Clark r | Scott Clark | 03m |
Claims Bayesian optimization outperforms grid and random search for hyperparameter tuning, reducing experiment time and cost; reports applicability to neural… | ||
| DBTB INT Konstantine Boudnik r | Konstantin Boudnik | 04m |
Describes a memory‑computing‑in‑the‑cloud engine called Maam Core used across customer deployments and argues that the Apache big data stack plus open source… | ||
| DBTB INT Prakash Chockalingam r | Prakash Chockalingam | 03m |
Practical checklist for running Spark Streaming pipelines on the Databricks platform with attention to production-scale concerns: how to handle state and… | ||
| DBTB INT Shubho Sengupta r | Shubho Sengupta | 02m |
Argues that large-scale supervised learning wins when data scale matches model size and gives specifics: convolutional nets with hundreds of millions of… | ||
| DBTB INT Dinesh Narayanan r | Dinesh Narayanan | 01m |
Samsung SSIC describes building distributed systems and deploying software in cloud environments, illustrating options for real-time and batch processing. The… | ||
| Justin Basilico interview | Justin Basilico | 04m |
Claims that combining high-quality data with solid intuition and algorithms produces results that solve previously intractable problems; building ML systems… | ||
| DBTB INT Oscar Celma r | Òscar Celma | 06m |
How Pandora balances playing familiar favorites with serendipitous discovery using real user signals and engineering checks is described concretely here… | ||
| Jeremiah O'Connor | Jeremiah O'Connor | 03m |
Research team builds a pipeline that crawls clear and dark web sources, applies unsupervised anomaly detection to find cybercrime patterns at internet scale… | ||
| DBTB INT Subbu Rama r | Subbu Rama | 03m |
Describes a working software to virtualize GPUs that pools heterogeneous accelerators in data centers and claims measurable gains: higher application… | ||
| DBTB INT Peter Bailis r | Peter Bailis | 02m |
Argues that collected data now outstrips human attention and claims the main bottleneck is prioritization rather than storage, so focus shifts to systems that… | ||
| DBTB INT Seema Jethani r | Seema Jethani | 02m |
Product claims that react kb and react es are production-ready distributed databases for putting government open data into hands that can query, deploy and… | ||
| DBTB INT Sandy Ryza r | Sandy Ryza | 04m |
Demonstrates a Spark-based time series library that focuses on concrete layout choices rather than algorithms, arguing that contiguous per-entity layout and… | ||
| DBTB INT Eric Anderson r | Eric Anderson | 03m |
Conference reflection argues that modern systems and abstractions now let a small team move beyond plumbing and deliver production data work much faster, but… | ||
| DBTB INT Nicholas Pilkington r | Nicholas Pilkington | 03m |
Claims that consumer drones are already practical for real business use and that off-the-shelf hardware yields high‑resolution, high‑cadence imagery now… | ||
| DBTB INT Anthony Goldbloom r | Anthony Goldbloom | 03m |
Kaggle runs public prediction challenges where thousands compete on real datasets and learning by doing beats lectures for many people, so competing in Kaggle… | ||
| DBTB INT Carlos Guestrin r | Carlos Guestrin | 04m |
Argues the field hits a phase transition from academic work to real-world intelligent apps, with winners like early Amazon recommender systems and Uber showing… | ||
| DBTB INT Gregory Stewart r | Gregory Stewart | 03m |
Algorithms are far more powerful than five years ago, pushing some problems into solvable territory while many real problems remain beyond current technical… | ||
| DBTB INT Anand Kulkarni r | Anand Kulkarni | 02m |
Claims that sales tasks formerly reserved for humans are now automatable using data and software and that LeadGenius applies those methods to real lead… | ||
| DBTB INT Andrew Or r | Andrew Or | 03m |
Overview of how Apache Spark hides its internal complexity: which runtime components (scheduler, shuffle, memory manager) enforce the abstractions, which APIs… | ||
| DBTB INT Yujay Huoh | Yujay Huoh | 03m |
Works at Earnest and frames mobile and sensor streams as ubiquitous, with phones generate constant data that teams already possess. Argues that modest… | ||
| DBTB INT S Srinivasan r | S. Srinivasan | 02m |
Argues that behavioral analytics must track individual attitude changes over time and that deriving that signal requires ingesting and processing very large… | ||
| DBTB INT Phillip Meisen r | Phillip Meisen | 02m |
Claims that treating time as a primary dimension multiplies the value of datasets, backed by 16 years of practical experience and examples like AlphaGo and… | ||
| DBTB INT Josh Wills r | Josh Wills | 03m |
Says working as director of data engineering at Slack and argues that tech advantage now often comes from a defensible data advantage held by companies like… | ||
| DBTB INT Lawrence Grodeska r | Lawrence Grodeska | 02m |
CivicMakers runs a San Francisco civic innovation studio that builds prototypes for agencies, nonprofits and civic startups and reports working directly with… | ||
| DBTB INT Jeremey Corbett r | Jeremey Corbett | 02m |
Built a production pipeline that ingests and processes a highly structured corpus and trains multiple specialized ML algorithms to solve real tasks, with… | ||
| DBTB INT Thomas Trolez r | Thomas Trolez | 02m |
Built a devreal.ai node linking talk, interview and Single Customer View pages and shows a production ETL that previously took up to 10 days now running in… | ||
| DBTB INT Jason Lally r | Jason Lally | 05m |
Program manager for Open Data at the City and County of San Francisco (Office of the Chief Data Officer) describes DataSF as a program that publishes city data… | ||
| Xavier Amatriain interview | Xavier Amatriain | 04m |
Claims that data and text are raw material that can be either noise or turned into usable knowledge, and that success is not volume but the pipeline and… | ||
| DBTB INT Richard Downe r | Richard Downe | 03m |
Casetext uses cheap, scalable compute to run older language algorithms blisteringly fast and ship an extractive summarization pipeline that pulls concise… | ||
| DBTB INT Ronald Sujithan r | Ronald Sujithan | 04m |
A company-built text analytics API service packages the messy parts of NLP so teams can skip building pipelines from scratch and focus on product use cases… | ||
| DBTB INT Daniel Lewis r | Daniel Lewis | 02m |
Claims that large bodies of case law are finally in machine-readable legal corpora and that value comes not from access to the same data but from asymmetric… | ||
| DBTB INT Morgan Keys r | Morgan Keys | 02m |
Built a set of data interfaces at OpenGov that parse messy ledgers and render local government accounting data into interactive views, developed over six to… | ||
| DBTB INT Alexis Roos r | Alexis Roos | 02m |
Argues that real-world tabular data often reveals much more when recast as a network, and demonstrates modeling and querying those relationships at scale using… | ||
| DBTB INT Ryan Orban r | Ryan Orban | 06m |
Galvanize runs immersive programs that teach coding for AI and data engineering and emphasizes data literacy and practical experience over theory, with the… | ||
| DBTB INT Alan Ngai r | Alan Ngai | 05m |
Claims that poor ops telemetry changes data and model outputs, and that assembling end-to-end monitoring across pipelines fixes it by letting teams organize… | ||
| DBTB INT Max Ho r | Max Ho | 02m |
Explains how Course Hero turns sprawling product telemetry into a short list of actionable insights, filters raw event logs into prioritized hypotheses, and… | ||
| DBTB INT Evion Kim r | Evion Kim | 05m |
Machine learning engineer at Mattermark argues that shipping models requires engineering capacity, not just model ideas, and recommends checking team… | ||
| DBTB INT Francoise Gilbert r | Francoise Gilbert | 02m |
A Greenberg Traurig attorney argues that codes of conduct plus greater awareness of ethical obligations and existing laws make big data work rather than stall… | ||
| DBTB INT Michael Terry r | Michael Terry | 03m |
Claims that modern tooling and data integrations make analytics accessible to most developers and business owners, not just big tech. Argues the same stacks… | ||
| DBTB INT Finton f | Finton | 03m |
Claims that KX time-series stack ports from finance to IoT and handles finance-scale volumes for device telemetry, showing a working prototype for sustained… | ||
| DBTB INT Gary Kazantsev | Gary Kazantsev | 04m |
Runs Bloomberg's machine learning engineering group and frames the field as the intersection of exponentially growing compute and slower-growing… | ||
| DBTB INT Antonio Valderrabanos | Antonio Valderrabanos | 02m |
This talk video is a node in devreal.ai, the developer community graph.
It connects to:
Talk: Aspect Based Sentiment Analysis in 20+ L... —… | ||
| DBTB INT Vita Markman | Vita Markman | 02m |
At LinkedIn the central claim is that machine learning is not magic and supervised models require clean and consistent data so expect practical rules not… | ||
| DBTB INT Greg Lindahl | Greg Lindahl | 04m |
Describes efforts to make Internet Archive collections more accessible, with 12 petabytes compressed of web archive back to 1996, plans to scan 10 million… | ||
| DBTB INT Xiao Fan | Xiao Fan | 02m |
Workday engineering describes an approach to improving search across internal products, framing search as a time‑saving utility rather than a research project… | ||
| DBTB INT Scott Waterman | Scott Waterman | 03m |
Argues that treating text as technology—not static documents but artifacts tied to computation—lets web-scale pipelines and search indexers reveal relations… | ||
| DBTB INT Jill Desmond | Jill Desmond | 03m |
Works at VersaMe as a senior data scientist and describes an iterative workflow that abandons optimistic first passes and rebuilds data collection until labels… | ||
| DBTB INT Michelle Casbon | Michelle Casbon | 04m |
A github repo for idml persistence layer is unveiled containing core ML libraries that aim to operationalize NLP by wiring common algorithms into a reusable… | ||
| DBTB INT Barak Turovsky | Barak Turovsky | 02m |
Google Translate is presented as one of the oldest and largest machine learning systems with a 10th anniversary milestone and a recent disruptive shift toward… | ||
| DBTB INT Maarten | Maarten Bosma | 03m |
Claims that more data is created in the last two years than in all previous history and uses that scale to power suggesting images to posts by mapping short… | ||
| DBTB INT Niyati Parameswaran f | Niyati Parameswaran | 04m |
Two lightweight frameworks get described and demonstrated: one for face detection and related generative work, the other for automatic text summarization; both… | ||
| DBTB INT David Talby | David Talby | 02m |
A short interview with David Talby and Alexy Khrabrov about their work at John Snow Labs and connections in the devreal.ai developer community graph. It… | ||
| DBTB INT Omar Alonso | Omar Alonso | 02m |
Principal data scientist at Microsoft describing how to improve model outcomes by focusing on good quality labels rather than cutting-edge models, outlining a… | ||
| DBTB INT Nemanja Spasojevic | Nemanja Spasojevic | 02m |
Data science leadership at Elysium Technologies reports that modern stacks let teams prototype analytics with low effort, focusing on slicing and dicing data… | ||
| DBTB INT Grant Ingersoll | Grant Ingersoll | 04m |
Lucidworks CTO argues that modern search moves beyond keywords to act as a ranking engine that surfaces what matters, recommending use of open source search… | ||
| DBTB INT Ricardo Baeza Yates | Ricardo Baeza-Yates | 03m |
This talk video is a node in devreal.ai, the developer community graph.
It connects to:
Talk: Data and Algorithmic Bias in the Web —… | ||
| DBTB INT Martin Froehler | Martin Froehler | 02m |
Claims machine learning works on financial data and links that claim to a quantitative finance pedigree and examples connected to Quantiacs. Describes concrete… | ||
| DBTB Roman Shaposhnik | Roman Shaposhnik | 05m |
Conference-level takeaways: the industry is in a peak of confusion about building big-data applications while data engineering and application development… | ||
| DBTB INT Shinji Kim | Shinji Kim | 05m |
Concord is a distributed stream-processing framework written in C++ and built on Apache Mesos that targets low-latency, multi-service real-time pipelines… | ||
| DBTB INT Hiral Patel | Hiral Patel | 05m |
Argues that open source tools let anyone build machine learning pipelines on a laptop without paying for enterprise software, and recommends prototype first to… | ||
| DBTB INT Chris Fregly | Chris Fregly | 05m |
Argues that for high-throughput streaming workloads exact answers cost too much and probabilistic approximations often meet production needs, demonstrates this… | ||
| DBTB INT Roland Tritsch | Roland Tritsch | 04m |
Vice president of engineering at Nitro describes concrete shifts: data engineering advances over the last five years unlock problems previously out of reach… | ||
| DBTB INT Slava Akhmechet | Slava Akhmechet | 05m |
RethinkDB evangelizes that the web is shifting from snapshots to live, real-time apps because modern sensors, storage and connectivity let systems ingest and… | ||
| DBTB INT Zack Loebel Begelman | Zack Loebel-Begelman | 04m |
Credit Karma analytics team deals with billion+ event datasets and tries to pipeline them with Akka tech but discovers a concrete failure: a homegrown… | ||
| DBTB INT Vladimir Bacvanski | Vladimir Bacvanski | 04m |
Claims a new unnamed technology delivers excellent performance and lets teams build big data pipelines more quickly, and recommends a concrete learning path… | ||
| DBTB INT Ilya Ganelin | Ilya Ganelin | 02m |
Describes deep work on streaming technologies and experiments with in-memory databases, claiming latency goals on the order of milliseconds. Emphasizes systems… | ||
| DBTB INT Monal Daxini | Monal Daxini | 03m |
Argues that real-time streaming data is ubiquitous and that a correct answer is building reusable pipelines rather than point analyses, describing work on a… | ||
| DBTB INT Manish Singh | Manish Singh | 03m |
CTO of MityLytics argues that generic benchmarks mislead and that real projects require run your own loads and export actual job traces to characterize… | ||
| DBTB INT Malcolm Greives | Malcolm Greives | 04m |
Asserts that functional programming is the clearest path to correctness, arguing there are "a million ways" to be right but FP makes the simplest correct… | ||
| DBTB INT Kanu Gulati | Kanu Gulati | 05m |
Argues that combining large volumes of data with learning algorithms, distributed cloud infrastructure and modern hardware ushers an intelligence computing era… | ||
| DBTB INT Shourabh Rawat | Shourabh Rawat | 04m |
Argues that image recognition can be deployed as an internal service by a small data team using the existing Python web stack, showing an architecture that… | ||
| DBTB INT Adam Mollenkopf | Adam Mollenkopf | 04m |
Claims that DC/OS gives a common operating environment that runs and reschedules services so teams stop managing machines, with Spark, Kafka, Cassandra… | ||
| Jay Kreps interview | Jay Kreps | 06m |
Argues that treating data as a continuously updated stream changes architecture and makes real‑time processing practical, and describes concrete pieces built… | ||
| DBTB INT Greg Rahn | Greg Rahn | 03m |
Argues that JSON is self-describing and ubiquitous and surveys the actual ways teams handle it: using relational SQL engines, document databases, or… | ||
| DBTB INT Sal Uryasev | Sal Uryasev | 04m |
GoFundMe donation data is an untapped signal that reveals who gives, how much, and which causes attract funds; adding third‑party sources consistently uncovers… | ||
| DBTB INT Parag Namjoshi | Parag Namjoshi | 03m |
Workday frames ten years of transaction logs as product-differentiating data and shows a practical shift from commoditized systems to using those logs for… | ||
| reactive James Roper 4 21 16 | James Roper | 01h 22m |
A sample social app called Chipper is refactored into microservices using Lagom and Play with SBT, a service locator and gateway, and Cassandra as the default… | ||
| Distributed Visualization for Genomic Analysis | Alyssa Morrow, Eric Tu | 23m |
Mango is a distributed genome browser that materializes requested regions from persistent storage (BAM/VCF or Apache Parquet via ADAM) into memory organized as… | ||
| How Data helps UX evolves at realtor.com | Ian Lin | 23m |
Realtor.com finds local neighborhood pages get almost no upstream traffic, with over 80% of users leaving within one minute and only one of the top 20… | ||
| Real Time Machine Learning Visualization with Spark | Chester Chen | 19m |
Adds callback hooks to Spark ML algorithms (K‑Means, logistic and linear regression) to emit iteration-specific messages including centroids, cost and a… | ||
| Towards a virtual reality meta-Earth | Bo Huang | 18m |
A working pipeline converts OpenStreetMap into a real‑time 3D meta‑Earth using the Unity game engine, Overpass API queries for region slices, and scientific… | ||
| Panel: Artificial Intelligence: Who is Winning?Panel d4 | Alexy Khrabrov, Pete Skomoroch, Richard Socher, Shivon Zillis, SriSatish Ambati | 59m |
Panel discussion weighs why big firms buy startups: a war for talent and for Salesforce acquisitions of Metamind and others, not magic models. Speakers… | ||
| Optimizing Machine Learning Models | Scott Clark | 41m |
Bayesian global optimization via SigOpt claims to find better hyperparameters far faster than grid or random search, trading exploration and exploitation using… | ||
| Fast deep recurrent net training | Sanjeev Satheesh | 25m |
Deep Speech is a convolutional-to-biRNN speech recognizer trained with CTC; a seven-layer bidirectional baseline gets stuck at a training glass ceiling… | ||
| Analyzing Massive Time Series Data with Spark | Sandy Ryza | 21m |
Open-source Spark-TS library built on Apache Spark and exposing Scala/Java/Python APIs implements a time series RDD that stores each key’s full series plus a… | ||
| What Kaggle has learned from 2MM machine learning models | Anthony Goldbloom | 41m |
Kaggle reports 2 million models submitted and now sees over 100,000 model submissions per month, each taking roughly 5–10 hours to develop, which yields… | ||
| Keynote: How Can We Trust Machine Learning? | Carlos Guestrin | 46m |
Argues a concrete path to trust: uses GraphLab Create and deployable microservices to take models from laptop to real-time AWS services, shows a deep network… | ||
| Building Machine Learning Pipeline Using Aerosolve | Peng Ye | 49m |
Aerosolve is an open source pipeline used at Airbnb to power Price Tips and Smart Pricing, training generalized additive models (piecewise linear splines +… | ||
| UX Techniques Supporting Varying Levels of Aggregation | Katherine Ahern | 42m |
Aggregation choice flips conclusions: the same dataset shows 48 of 50 states example where more states’ averages fall yet the population-weighted average… | ||
| Build Smarter, Don't Just Build | Hernan Asorey, Robin Glinton | 22m |
A four-person product data science team scales insights across dozens of clouds by automating three bottlenecks: cataloging raw logs with Elation to cut… | ||
| Challenges Applying Traditional UI/UX Principles to M.L. | Demetri Spanos | 39m |
A real-time transaction classifier flags a supposedly impossible column value (50 where schema allows 1–10), which turns out to be an upstream data change and… | ||
| Visualizing the silent force of the Bay | Boriana Viljoen | 21m |
Kiteboard currents in the San Francisco Bay behave like a violent river and conventional tidal charts often fail to warn users; a user-centered study of 110… | ||
| Designing Visualizations of Health that Resonate With Users | Alan McLean | 21m |
Argues that health visuals must be data-first and user-tested for mobile, using Fitbit’s Reminders to Move as a case study; sketches are throwaway prototypes… | ||
| Translating a Trillion Pointo Diseases of Data into Therapies | Atul Butte | 20m |
Lab converts public genomic and clinical repositories into finished products: a hemopexin blood test for preeclampsia derived from meta-analysis of hundreds of… | ||
| Visualizing big data | Olivier Balbous | 41m |
Builds an interactive JavaScript dataviz connected to a couple hundred million rows in Amazon Redshift, using D3.js for visuals and OpenBouquet as the… | ||
| Pro-Active: Designing a genuinely helpful SQL interface | Aaron Kalb | 39m |
A proactive SQL interface that uses semantic search + fuzzy matching and ranks suggestions by social statistics to surface the most-used, fresher tables and… | ||
| Remixing Design with Data | James Mulholland | 21m |
Platfora describes a data-driven design process: use the Kano model survey to score features by avoided frustration versus added delight, then prioritize the… | ||
| When Visualization Best Practices Fall On Deaf Ears | Akash Mukherjee | 22m |
Real-world dashboards at Facebook force tradeoffs between academic encodings and business needs: the claim is that in practice you must increase click-through… | ||
| Protecting data scientists in healthcare with type safety | Gopal Ramachandran | 22m |
Wellframe moves ingestion out of ad hoc Python into a Scala pipeline that puts Kafka between sources and consumers and emits Scala case classes for strong QA… | ||
| The Lego Model for Machine Learning | Vitaly Gordon | 42m |
A machine learning platform called Optimus Prime built on Spark + Scala generates code to automate a ten‑step feature engineering pipeline: feature typing… | ||
| Extracting Medical Attributes from Clinical Trials | Sanghamitra Deb | 26m |
Extracts diseases, dosages, ages, genders and side effects from FDA drug labels and clinical trials using noun/phrase candidates, word2vec/Gensim, the Stanford… | ||
| What Healthcare Can Learn from Netflix | Eric Williams | 19m |
Omada Health builds personalization with analytics, machine learning and built‑in experimentation across 60,000 participants, mailing physical scales and… | ||
| With Big Data Comes Big Responsibility | Nish Bhat | 20m |
Genetic testing moves from niche to scale: cost drops faster than Moore’s Law and consumer tests move from $250 test vs $4,000 in price. Industry needs to stop… | ||
| Why you can't afford to ignore deep learning | Jeremy Howard | 43m |
Argues deep learning already outperforms humans on vision and speech tasks, cites a state-of-the-art algorithm for early detection of lung cancer built with… | ||
| Panel: Living Data | Alexy Khrabrov, Daniella Perlroth, Jeff Lerman, John St. John, priya joseph | 01h 02m |
This talk video is a node in devreal.ai, the developer community graph.
It connects to:
Talk: Panel: Living Data —… | ||
| Data Design Challenges for Enterprise IoT Applications | Zachary Taschdjian | 21m |
Claim: prioritize a data-first design process because enterprise IoT data breaks common design assumptions. Concrete issues: fleets of 40–50,000 assets with… | ||
| Interactive Machine Learning on Genomics Data with Spark Notebook | Andy Petrella | 21m |
Converts 1000 Genomes BAM/VCF raw files (≈152 GB) into ADAM schema stored as Parquet, creating ~10,000 partitions so the same dataset loads in about one minute… | ||
| A Scalable GA4GH Server Implementation | Andy Petrella | 24m |
A GA4GH server implementation that claims to be a scalable reference uses Apache Spark and the ADAM genomics schemas as its back end, implements variant… | ||
| How Data Science is Evolving Technology Education | Ryan Orban | 25m |
Galvanize runs three immersive programs: a three to six month full stack course, a 12 week data science bootcamp and new data engineering training and reports… | ||
| Ginger.io: a Data-Driven Mental Health Care Provider | Sai Moturu | 18m |
Ginger.io delivers mental health care through a smartphone app that combines passive behavioral sensing (movement, communication, sleep) with 1.3M self‑reports… | ||
| Let's Build Civic Tech to Scale Democracy for a Better Future | Matt Mahan | 31m |
Argues that US democracy falters because voting and representation fail to scale with population and complexity, citing 40 elected officials per voter, over… | ||
| Building realtime efficient queries in an IOT platform | Dinesh Narayanan | 20m |
An IoT pipeline that ingests via WebSocket/CoAP uses Avro/MessagePack, Kafka + Schema Registry and converts rows to Apache Parquet for HDFS, yielding about… | ||
| Active Learning and Human-in-the-Loop Machine Learning | Lukas Biewald | 30m |
Argues that model gains come more from larger and cleaner labeled datasets than from tweaking algorithms, showing a toy result where error falls from ~13% at… | ||
| Creating Value by Turning Law Into Data | Daniel Lewis | 23m |
Ravel Law digitizes court opinions (partnering with Harvard to convert 45,000 books) and applies cloud, Spark, machine learning and NLP to extract structured… | ||
| The practice of acquiring good labels | Omar Alonso | 39m |
Focuses on concrete methods: program crowds with clear instructions and short hits, use two-phase filtering (cheap Mechanical Turk binary filters then… | ||
| Open Source Memory Speed Virtual Distributed Storage System | Gene Pang | 40m |
Alluxio (formerly Tachyon) is an open‑source, memory‑first virtual distributed storage layer that sits between compute frameworks (Spark, Flink) and backend… | ||
| Twitter Heron in Practice | Karthik Ramasamy | 39m |
Heron runs in production at Twitter for nearly two years, processing billions of events and replacing Storm with a Storm-compatible topology model that runs as… | ||
| UX and the Changing Relationship Between Mind and Machine | Hunter Whitney | 43m |
Argues that machine learning and deep learning are tightening feedback loops with people, using concrete examples: email response suggestions, fraud-detection… | ||
| Host Sponsor Welcome: IBM Analytics | Joel S. Horwitz | 08m |
IBM announces a year-old investment in Apache Spark and positions it as an analytics operating system, citing internal organic adoption rather than a mandate… | ||
| Importance of rethinking data visualization | Kyle Hailey | 19m |
Argues that data visualization needs domain expertise and shows specific fixes: redoes NASA Challenger charts with a Tufte replot to reveal a… | ||
| Host Sponsor Welcome: Nitro | Roland Tritsch | 07m |
A career narrative that lists concrete systems and outcomes: a client‑server bill‑of‑materials built for General Motors on an early relational database; large… | ||
| How to Use Big Data to Inspire Consumer Confidence | Daniel Doerr | 22m |
Credit Karma uses data from 50 million members and machine learning to turn credit reports and account links into a quantitative, composite measure of personal… | ||
| Like2Vec with neural network embeddings | Mike Tamir, Marvin Bertin, David Ott | 37m |
Applies Word2Vec-style shallow neural embeddings to user–item graphs, building Like2Vec embeddings from an item co-occurrence matrix weighted by log‑likelihood… | ||
| Cancer Screening using Deep Learning | Daniel Golden | 18m |
Arterys automates cardiac MRI segmentation on ssfp scans by running convolutional nets in the cloud and handling partially labeled clinical data. The pipeline… | ||
| From text to knowledge via ML algorithms - the Quora answer | Xavier Amatriain | 38m |
Quora claims over 80 million monthly visitors and uses machine learning to turn short Q&A text into actionable signals: a personalized learning-to-rank… | ||
| On the path to data Nirvana: Supporting OLTP and OLAP on Hadoop | Monte Zweben | 38m |
Claims a single, scale‑out RDBMS called Splice Machine runs OLTP and OLAP side‑by‑side by using HBase for OLTP and Spark for OLAP, with a cost‑based optimizer… | ||
| nothing matches | ||