AI-powered GPU kernel optimization for NVIDIA & AMD, zero code changes required
Try the live demo below to see Mako in action
running on
running on
How It Works
The Mako Compiler and software stack lets you deploy AI like the pros, leveraging all the state of the art optimization tricks that the largest AI labs in the world use
Seamless Integration
Works out-of-the-box with any PyTorch or Hugging Face model – no code changes required
Automatic Optimization
Mako Compiler auto-tunes GPU kernels and inference engines to maximize performance
Continuous Improvement
Our search-based optimization engine keeps learning over time, making your AI faster and more efficient, 24/7
Performance
State of the Art Results
Mako optimized containers make deploying state-of-the-art inference optimizations on new hardware easy.
- Small-ish Models
- Mistral-7B achieves 49% performance improvement and 70% cost reduction.
- Long Context Models
- Phi-medium-128k achieves 31% cost reduction, but raw tokens per second declines.
- Larger models
- Qwen2-72B achieves an 85% performance improvement and 44% cost reduction
Meet the Team
Over 30 years of experience bringing high performance AI from research into production.
Waleed Atallah
Chief Executive Officer
Lukasz Dudziak
Chief Technology Officer
Mohamed Abdelfattah, Ph.D
Chief Science Officer
Angels and Advisors
Jeff Dean
Chief Scientist at Google
Micah Villmow
Fmr. Principal SWE at NVIDIA, TensorRT
David Levy
Fmr. Startups at AWS
Abdel-rahman Mohamed
Head of AI at Rembrand
Nicholas Lane
Prof. at Cambridge University and CSO at Flower
Ben Fielding
Co-founder and CEO at Gensyn
Recent Blogs
Shape the Future of AI with Us!
Join MAKO, where creativity meets cutting-edge innovation. Be part of a passionate team making breakthroughs in that shapes industries and changes lives.
Ready to make an impact?Explore Opportunities
Stay in Touch
Keep up with the latest on AI portability and performance.