Baris Kasikci
Associate Professor
Builds efficient and reliable computer systems, including serving and quantization systems for large language models.
Baris Kasikci's lab at UW works on efficient training & inference and large language models. 40 papers since 2023, 20 at top venues, mostly MLSys and ICLR. 9 PhD students. 13 alumni, most often to Google and Meta. Applications close .
At a glance
- papers, 2023–2026
- 10/ year
- at T1 venues
- 50%
- PhD students
- 9
- MS students
- 0
- alumni listed
- 13
Papers by yearfrom the lab homepage, 2023–2026
- 14 papers2026 · T1 8 · collecting
- 11 papers2025 · T1 5
- 10 papers2024 · T1 5
- 5 papers2023 · T1 2
T1 venuesOtherVenues: MLSys, ICLR, ICML, SOSP, EMNLP
Research focuspaper titles that use each topic's words
- Efficient training & inference11 papers
- Large language models10 papers
- quantization1 paper
- builds efficient0 papers
- computer systems0 papers
Paper list
From the lab homepage.
202614 papers
- MLSys (T1)Accelerating Large-Scale Reasoning Model Inference with Sparse Self-Speculative Decoding (opens in a new tab)
- International Conference on Architectural Support for Programming Languages and Operating SystemsBlendServe: Optimizing Offline Inference with Resource-Aware Batching (opens in a new tab)
- Conference on Language ModelingConsumerBench: Benchmarking Generative AI Applications on End-User Devices (opens in a new tab)
- Symposium on Operating Systems Design and ImplementationDiagnosing Performance Issues in Application-Defined Resources (opens in a new tab)
- MLSys (T1)DynaFlow: Transparent and Flexible Intra-Device Parallelism via Programmable Operator Scheduling (opens in a new tab)
- ICML (T1)Ekka: Automated Diagnosis of Silent Errors in LLM Inference (opens in a new tab)
- Conference on Programming Language Design and ImplementationHayroll: A Modular Wrapper for Translating C Macros and Conditional Compilation to Rust (opens in a new tab)
- Symposium on Operating Systems Design and ImplementationJul 2026
- NeurIPS (T1)M*: A Modular, Extensible, Serving System for Multimodal Models (opens in a new tab)
- EuroSys (T1)PASS: A Power Adaptive Storage Server (opens in a new tab)
- MLSys (T1)RagInfer: Efficient Retrieval-Augmented Generation Inference with Lookahead Retrieval (opens in a new tab)
- ICLR (T1)Tactic: Adaptive Sparse Attention with Clustering and Distribution Fitting for Long-Context LLMs (opens in a new tab)
202511 papers
- Workshop on Hot Topics in Operating SystemsApiary: An OS for the Modern FPGA (opens in a new tab)
- ICLR (T1)Fiddler: CPU-GPU Orchestration for Fast Inference of Mixture-of-Experts Models (opens in a new tab)
- MLSys (T1)FlashInfer: Efficient and Customizable Attention Engine for LLM Inference Serving (opens in a new tab)
- International Symposium on High-Performance Computer ArchitectureFrom Optimal to Practical: Efficient Micro-op Cache Replacement Policies for Data Center Applications (opens in a new tab)
- NDSS (T1)GadgetMeter: Quantitatively and Accurately Gauging the Exploitability of Speculative Gadgets (opens in a new tab)
- EMNLP (T1)LiteASR: Efficient Automatic Speech Recognition with Low-Rank Approximation (opens in a new tab)
- SOSP (T1)Mitigating Application Resource Overload with Targeted Task Cancellation (opens in a new tab)
- Symposium on Operating Systems Design and ImplementationNanoFlow: Towards Optimal Large Language Model Serving Throughput (opens in a new tab)
- International Symposium on High-Performance Computer ArchitecturePalermo: Improving the Performance of Oblivious Memory using Protocol-Hardware Co-Design (opens in a new tab)
- ACM SIGPLAN Conference on Object-Oriented Programming, Systems, Languages, and ApplicationsScalable and Accurate Application-Level Crash-Consistency Testing via Representative Testing (opens in a new tab)
- ACM Symposium on Cloud ComputingSnap & Replay: A New Way to Analyze uarch-Scale Performance Bottlenecks for ML Accelerators (opens in a new tab)
202410 papers
- MLSys (T1)Atom: Low-bit Quantization for Efficient and Accurate LLM Serving (opens in a new tab)
- MICRO (T1)Beehive: A Flexible Network Stack for Direct-Attached Accelerators (opens in a new tab)
- SIGCOMM (T1)Best Student Paper Award
- Workshop on Hot Topics in Storage and File SystemsCan Storage Devices Be Power Adaptive? (opens in a new tab)
- International Conference on Architectural Support for Programming Languages and Operating SystemsLimoncello: Prefetchers for Scale (opens in a new tab)
- International Conference on Architectural Support for Programming Languages and Operating SystemsProactive Runtime Detection of Aging-Related Silent Data Corruptions: A Bottom-Up Approach (opens in a new tab)
- ICML (T1)Quest: Query-Aware Sparsity for Efficient Long-Context LLM Inference (opens in a new tab)
- International Conference on Architectural Support for Programming Languages and Operating SystemsRPG^2: Robust Profile-Guided Runtime Prefetch Generation (opens in a new tab)
- ISCA (T1)UDP: Utility-Driven Fetch Directed Instruction Prefetching (opens in a new tab)
- IEEE/IFIP International Conference on Dependable Systems and NetworksZipChannel: Cache Side-Channel Vulnerabilities in Compression Algorithms (opens in a new tab)
20235 papers
- International Conference on Embedded SoftwareCrossTalk: Making Low-Latency Fault Tolerance Cheap by Exploiting Redundant Networks (opens in a new tab)
- ACM SIGOPS Operating Systems ReviewOne Profile Fits All: Profile-Guided Linux Kernel Optimizations for Data Center Applications (opens in a new tab)
- S&P (T1)PCspooF: Compromising the Safety of Time-Triggered Ethernet (opens in a new tab)
- SOSP (T1)Siloz: Leveraging DRAM Subarray Groups to Prevent Inter-VM Rowhammer (opens in a new tab)
- International Conference on Architectural Support for Programming Languages and Operating SystemsVidi: Record Replay for Reconfigurable Hardware (opens in a new tab)
Where alumni went13 listed, 44% to industry
Industry 4Academia 5
Most often to
- Meta
- Oxide Computer Company
Also to
- Boston University
- Carnegie Mellon University
- Columbia University
Apply to UWFall 2027
UW Allen School Computer Science PhD
- Round
- Single round
- Opens
- GRE
- not required
- English
- See program page
- Fee
- See program page
Not yet confirmed for this intake.
Similar labsnearest by research summary
- CNSJTU
Zeyu Mi
- Efficient training & inference
- CNSJTU
Erhu Feng
- Efficient training & inference
- CHETH Zurich
Christian Holz
- Interfaces & interaction techniques
- SGNUS
Bingsheng He
- Accelerators, chips & architecture
- USNYU
Hongyi Wen
- Recommendation, search & web
- CNZhejiang
Xiaohang Wang
- Accelerators, chips & architecture
Your notes
Save this lab to keep notes.
Data
Checked against the lab homepage.
- Collected
- Verified
- Papers
- The lab homepage
- Members
- Lab homepage
- Alumni
- Lab homepage
Tiers are venue classifications, not lab rankings.
UWcloses