Speaker

Lin Zhao

Staff Engineer · Streamscale

StreamingKafka

Lin works on exactly-once streaming systems at Streamscale, running pipelines that move millions of events per second with strict ordering and replay guarantees.

Talks

Data Engineering

Exactly-Once at a Million Events per Second

Lessons from running mission-critical streaming pipelines with strict ordering and replay guarantees.

Chicago

Jul 17, 2024 · 3:00 PM

AI / ML

Real-Time Feature Stores Without the Pain

Bridging the online/offline gap with a single source of truth for features and labels.

Atlanta

Jul 12, 2022 · 10:30 AM

Published papers

Windowing Strategies for Late-Arriving Data in High-Throughput Streams

L. Zhao, K. Nakamura

Streaming · 2025 · 21 citations

Deterministic Replay in Exactly-Once Stream Processors

L. Zhao, K. Nakamura

Streaming · 2024 · 51 citations

Adaptive Checkpointing for Exactly-Once Stream Processing on Spot Instances

L. Zhao, B. Andersen

Streaming · 2024 · 26 citations

Unified Batch and Stream Processing with Stateful DoFn Abstractions

L. Zhao, Q. Zhang

Streaming · 2024 · 36 citations

GPU-Accelerated Window Aggregations on High-Velocity Streams

F. Almeida, L. Zhao

Streaming · 2023 · 52 citations

Online/Offline Consistency in Real-Time Feature Stores

L. Zhao, V. Cruz, R. Mehta

ML Platforms · 2022 · 118 citations

Consistent Hashing for Stateful Stream Operators Under Elastic Scaling

K. Nakamura, L. Zhao

Streaming · 2020 · 142 citations

Schema Evolution Strategies for Open Table Formats

A. Okafor, L. Zhao

Storage · 2019 · 271 citations