Databricks Takes 2nd in Meta Muse Spark Kernel Challenge
Autonomous agent loops are now producing production-grade GPU kernels that beat mainstream inference-framework baselines, which lowers the cost floor for serving AI workloads. Engineering teams running high-volume inference can treat LLM-driven kernel optimization as a practical pipeline rather than a research demo. The finding that Opus 5.5 outperforms other models on kernel optimization also gives practitioners a concrete model choice for similar code-optimization agents.
Databricks placed 2nd out of 300,000+ submissions in Meta's Muse Spark Open Kernel Challenge, with its GPU kernels delivering an average 2.5X speedup over SGLang kernels.
The team built a Kernel Factory in which AI agents autonomously hill-climb toward better kernels, iteratively testing and refining GPU code without human-guided search at each step. The company found Claude Opus 5.5 to be the strongest model for GPU kernel optimization among the models it evaluated.
The kernels are not just competition artifacts: Databricks deploys them directly into production, where they power Unity Gateway and AI inference workloads for its customers.
Databricks just took 2nd place in Meta's Muse Spark Open Kernel Challenge, out of 300K+ submissions!
Our kernels achieved an average 2.5X speedup over SGLang kernels.
We built a Kernel Factory where AI agents autonomously hill-climb their way to better kernels. We found Opus 5.5 to be the strongest model for GPU kernel optimization.
These kernels go straight from our Kernel Factory into production, powering Databricks Unity Gateway and AI inference for our customers.
This report is categorized under AITOP's "Industry" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Databricks, Claude Opus 5.5, SGLang, agents.
Source attribution: curated from Yuchen Jin (@Yuchenj_UW).
