Wei Wang
where else to find them
linkedin:weiww
what they talk about
all 18 in the filters
where they keep going back
all 22 in the filters
on a stage with
all 132 in the filters
who talks about the same things
- Xiaoya Xia8 talks
- Willem Jiang16 talks
- Edidiong Asikpo8 talks
- Sujata Tibrewala4 talks
- Chris DiBona8 talks
- Kevin Lee11 talks
| title | conference | date |
|---|---|---|
| The State of the PyTorch Ecosystem in 2026: Global Trends and China’s Rising Role | KubeCon + CloudNativeCon + OpenInfra Summit + PyTorch Conference China 2026 | 2026-09-07 |
In 2026, PyTorch stands at the center of a rapidly expanding global AI ecosystem. Through a data-driven analysis of activity across GitHub, Hugging Face, and… | ||
| RollArt: Disaggregated Multi-Task Agentic RL Training at Scale | 20th USENIX Symposium on Operating Systems Design and Implementation (OSDI '26) | 2026-07-13 |
RollArt disaggregates multi-task agentic RL by mapping prefill tasks to compute-optimized GPUs, decode tasks to bandwidth-optimized GPUs, and environments to… | ||
| Attack of the Bubbles: Straggler-Resilient Pipeline Parallelism for Large Model Training | NSDI '26 | 2026-05-04 |
PIPEMORPH is a system that adapts pipeline schedules and offloads communication to host memory with CPU-side RDMA delegation to eliminate head-of-line blocking… | ||
| RollPacker: Taming Long-Tail Rollouts for RL Post-Training with Tail Batching | NSDI '26 | 2026-05-04 |
RollPacker implements tail batching: it groups prompts that produce very long responses into occasional long rounds and uses speculative execution to run extra… | ||
| RollingEvidence: Autoregressive Video Evidence via Rolling Shutter Effect | USENIX Security '25 | 2025-08-13 |
An active system embeds rolling shutter stripes by modulating an LED with a 16‑FSK LED probe and records them on commodity CMOS phones; an autoregressive… | ||
| Torpor: GPU-Enabled Serverless Computing for Low-Latency, Resource-Efficient... | USENIX ATC '25 | 2025-07-07 |
Torpor implements late binding with model swapping to keep models in host memory and move them to a shared GPU pool on demand, using asynchronous API… | ||
| Toppings: CPU-Assisted, Rank-Aware Adapter Serving for LLM Inference | USENIX ATC '25 | 2025-07-07 |
System pins the base LLM on GPUs and dynamically loads LoRA adapters from host memory, while CPU-assisted LoRA computation precomputes the lightweight XAB… | ||
| GREYHOUND: Hunting Fail-Slows in Hybrid-Parallel Training at Scale | USENIX ATC '25 | 2025-07-07 |
Characterization on a 10,000+ GPU multi-tenant cluster finds transient fail-slows from CPU contention, GPU thermal throttling, and network congestion that last… | ||
| GPU-Disaggregated Serving for Deep Learning Recommendation Models at Scale | NSDI '25 | 2025-04-28 |
Prism disaggregates CPU and GPU pools via graph-level partitioning and RDMA so embedding lookups run on CPU nodes and dense nets on GPU nodes, using a fusgraph… | ||
| AutoSketch: Automatic Sketch-Oriented Compiler for Query-driven Network Telemetry | 21st USENIX Symposium on Networked Systems Design and Implementation (NSDI '24) | 2024-04-16 |
AutoSketch automatically compiles high-level query operators into sketch instances, then searches configurations to meet a user-specified accuracy intent… | ||
| Beware of Fragmentation: Scheduling GPU-Sharing Workloads with Fragmentation... | USENIX ATC '23 | 2023-07-10 |
Paper defines a statistical fragmentation metric that predicts expected unallocatable GPU capacity given task distributions, implements a scheduler called… | ||
| FLASH: Towards a High-performance Hardware Acceleration Architecture for Cross-silo... | NSDI 2023 | 2023-04-17 |
Identifies nine real-world cryptographic operations that dominate cross-silo federated learning and shows they all reduce to two primitives, modular… | ||
| Following the Data, Not the Function: Rethinking Function Orchestration in Serverless | NSDI 2023 | 2023-04-17 |
Pheromone implements a data-bucket abstraction and trigger primitives on Kubernetes (coordinator + workers) so intermediate objects themselves drive invocation… | ||
| TCP memory isolation on multi-tenant servers | Linux Plumbers Conference 2022 | 2022-09-12 |
Proposes removing the global tcp_mem limit and charging TCP socket buffers to cgroup memory limits, implements a measurement mode to gather per-job TCP memory… | ||
| Panel: A Decade of Open Source in China - Insights Report with Chinese OSPO's Best Practices | Open Source Summit North America 2022 | 2022-06-21 |
Chinese firms like Alibaba report more than 3,000 GitHub repositories and over one million stars, while xlab analyzes 86 million GitHub event logs in 2020 and… | ||
| MLaaS in the Wild: Workload Analysis and Scheduling in Large-Scale Heterogeneous GPU... | 19th USENIX Symposium on Networked Systems Design and Implementation (NSDI '22) | 2022-04-04 |
A two-month Alibaba MLaaS trace covers over 6,700 GPUs across ~1,800 nodes with 1.2M tasks, 7.5M instances and shows concrete pain points: 85% gang-scheduled… | ||
| SonicASL: An Acoustic-based Sign Language Gesture Recognizer Using Earphones | UbiComp 2021 | 2021-09-21 |
A wearable system uses commodity earphones with a front-facing speaker and microphone to emit ultrasound and capture reflections, sending the audio to a phone… | ||
| Demystifying the Vetting Process of Voice-controlled Skills on Markets | UbiComp 2021 | 2021-09-21 |
Paper reverse-engineers skill vetting and shows concrete weaknesses: a relevance-based Turing test using ConceptNet distinguishes human vs machine vetters… | ||
| Peer-to-Peer Localization for Single-Antenna Devices | UbiComp 2021 | 2021-09-21 |
P2PLocate localizes a single-antenna device by co-locating a batteryless on-body backscatter tag and using the tag to create intentional multipath plus user… | ||
| TDX Live Migration | KVM Forum 2021 | 2021-09-15 |
Implements live migration for Intel TDX by adding a microTD helper that establishes a secure migration key and uses TD module sim calls to export/import… | ||
| CrystalPerf: Learning to Characterize the Performance of Dataflow Computation... | USENIX ATC '21 | 2021-07-14 |
CrystalPerf learns from source code to infer per-operator resource breakdowns (CPU, network, disk) and combines that with non‑intrusive sampling profiler… | ||
| Verification and Redesign of OFDM Backscatter | NSDI ’21 | 2021-04-12 |
Identifies why OFDM backscatter fails: WiFi receiver phase error correction cancels symbol-level tag phase shifts, breaking codeword-translation designs… | ||
| Work-in-Progress Reports (WiPs) | FAST '21 | 2021-02-23 |
Seven short systems results: an application‑local user‑space library filesystem for persistent memory shares VM pages, omits ACLs, and in single‑process… | ||
| Live Migration With Hardware Acceleration | KVM Forum 2020 | 2020-10-28 |
Offloads VM memory compression to a PCIe compression accelerator (Intel QAT) and plans to add DSA/DLA/IAA; implements zero copy DMA to guest memory, multi-page… | ||
| BatchCrypt: Efficient Homomorphic Encryption for Cross-Silo Federated Learning | USENIX ATC '20 | 2020-07-15 |
BatchCrypt encodes batches of quantized gradients into single long integers, uses two's-complement packing plus zeros padding and an analytical clipping method… | ||
| Retwork: Exploring Reader Network with COTS RFID Systems | USENIX ATC '20 | 2020-07-15 |
Protocol identifies reader contention by piggybacking tag inventory flags to mimic a one-to-many write under EPC Gen2, then runs selective reads to mark… | ||
| VMscatter: A Versatile MIMO Backscatter | NSDI 2020 | 2020-02-25 |
Implements space-time coding on tag using only on/off reflection states to achieve full MIMO diversity, with a practical hardware prototype (tested as 2×4×4)… | ||
| MArk: Exploiting Cloud Services for Cost-Effective, SLO-Aware Machine Learning Inference Serving | USENIX ATC '19 | 2019-07-10 |
MArk is an AWS inference-serving system that combines dynamic batching (to make GPUs cheaper than CPUs at moderate batch sizes), predictive autoscaling (to… | ||
| Open Source Education: Take CNCF Projects as an Example | KubeCon + CloudNativeCon China 2019 | 2019-06-24 |
Claims a campus platform that maps students to CNCF projects and provides one-click course environment with ready Kubernetes instances, Jupyter notebooks and… | ||
| Open Source Education: Take CNCF Projects as an Example | KubeCon + CloudNativeCon + Open Source Summit China 2019 | 2019-06-24 |
University builds a one-stop practical training platform that spins up CNCF stacks like Kubernetes, Envoy and containerd so students run real labs with a… | ||
| Updating your apps for location permission changes in Android Q | Google I/O 2019 | 2019-05-07 |
Android Q introduces concrete changes: apps can request allow only while using the app via ACCESS_FINE/COARSE mapped to ACCESS_BACKGROUND_LOCATION only when… | ||
| nothing matches | ||