Merge, Makora partner on autonomous GPU optimization
Merge has announced a partnership with Makora AI, whose autonomous systems automatically write, optimize, and deploy GPU code across the full inference stack. By automating performance tuning from high-level routing down to custom CUDA and Triton kernels, Makora replaces the slow, manual effort of optimizing AI workloads one model, chip, and kernel at a time across heterogeneous hardware.
Automating low-level GPU optimization across routing and custom kernels is essential for eliminating engineering bottlenecks as hardware ecosystems diversify.
- –Manual CUDA and Triton kernel tuning per model and chip is a major engineering bottleneck for AI teams.
- –Autonomous stack-wide optimization enables continuous performance improvements across heterogeneous GPU platforms.
- –Deep integration between infrastructure providers and agentic compiler tools signals a shift toward self-tuning compute stacks.
DISCOVERED
46d ago
2026-08-06
PUBLISHED
46d ago
2026-08-06
RELEVANCE
AUTHOR
merge_api