Data @Scale Spring 2022
as the organisers put it
Data @Scale is a technical conference for engineers who are interested in building, operating, and using data systems at scale. Data already enables companies to build products with user empathy, find new market opportunities, understand trends, make better decisions, and ensure that their services and systems stay healthy. The landscape of data systems is quickly […]
who it's for
machine learning engineers, data engineers, ML platform engineers, MLOps engineers, infrastructure engineers, research scientists
topics
self-supervised learning pretraining; energy-based models; Transformers as associative memory; Ray tasks and distributed object store; streaming feature joins for ML; incremental retraining pipelines; real-time feature processing platforms; TorchData and Arrow DataFrame; S3-backed Pub/Sub for ingestion; real-time SQL over JSON (converged index); model performance monitoring and explainability; typed workflow orchestration and schema testing
| title | author | length |
|---|---|---|
| Fireside Chat with Aparna Ramani & Yann LeCun | Aparna Ramani, Yann LeCun | 31m |
Argues that self-supervised learning — “fill in the blanks” pretraining becomes the default: massive unlabeled pretraining (example: ~a billion Instagram… | ||
| Keynote – Ion Stoica | Ion Stoica | 20m |
Claims that Ray consolidates streaming, training, tuning and serving into one system and replaces stitching disparate frameworks, with a minimalist six-call… | ||
| Keynote – Aparna Ramani | Aparna Ramani | 17m |
Meta describes a deep redesign of data infra to serve ML at scale including rack designs that link thousands of GPUs, new liquid cooling and faster networks… | ||
| ML Monitoring & Observability | Partha Kanuparthy, Animesh Dalakoti, Kunal Bhalla | 26m |
Meta diagnoses spikes in normalized entropy (any) explosions by correlating training logs, feature stats and lineage, and automates fixes like hyperparameter… | ||
| Automated Model Update & Evaluation | Chip Huyen | 29m |
Breaks down when to use stateful (incremental) vs stateless retraining and reports a real example: switching to incremental retraining cuts training cost ~45x… | ||
| Real-Time Data Processing for ML Feature Engineering | Weiran Liu, Ping Chen | 14m |
Meta describes a consolidated real‑time platform that evolved from Stylus, Puma, Turbine into an Extreme platform with an engine that emits machine code or… | ||
| Data Preprocessing for ML at Production Scale | Wenlei Xie, Vitaly Fedyunin, Yingxin Kang | 18m |
Describes a production pipeline that replaces Python list dict hacks with a streaming, composable DataPipe system and an Arrow-backed DataFrame to scale to… | ||
| Scalable Data Transportation & Ingestion | Ambud Sharma | 12m |
MemQ is a Pub/Sub that offloads replication to object storage and stores data as packetized batches so ML training can ingest petabytes cheaply. It uses S3 as… | ||
| Enabling Machine Learning through Real-Time Data Processing | Dhruba Borthakur | 20m |
Rockset is a real‑time indexing database built on RocksDB that claims millisecond SQL queries over JSON/Parquet without schematization by converting incoming… | ||
| Minimize Risks and Accelerate MLOps with Model Performance Monitoring and Explainability | Krishna Gade, Krishna Gade Fiddler | 19m |
Fiddler ships a model performance management system that compares production and training distributions using Jensen‑Shannon divergence to flag prediction… | ||
| Industrial-Scale Machine Learning | Andrea Olgiati | 20m |
AWS claims customers run over a million labeling tasks per day and serve 100 billion+ predictions monthly while some deploy millions of personalized models, so… | ||
| Making Data Quality an integral part of developing Machine Learning and Data Products | Ketan Umare, Katrina Rogan | 14m |
Describes a concrete pipeline for California housing: a type-safe orchestration platform that wires typed Python tasks into reproducible workflows, plus a… | ||
| nothing matches | ||