Gigantor Launches the Next Evolution of AI Acceleration
April 19, 2021 at 09:39 AM EDT
Gigantor Technologies Inc. is set to disrupt the market with its multiple patent-pending revolutionary technology, GigaMACS. Co-founders Don Gaspar, CEO, and Mark Mathews, CTO, formed Gigantor after analyzing the market’s existing approach to acceleration using stock hardware solutions or imprecise model altering strategies.
GigaMACS optimizes your machine learning model by converting it to a synchronal parallel circuit that runs as fast as the input pixel rate while minimizing latency. Most importantly, GigaMACS does not require changes to the model, and it will accelerate your convolutional neural network model to new heights.
“The key to maximum performance is to take the exact computation of your model and express it as an optimized, zero overhead pipelined circuit by eliminating redundancy and bottlenecks from your model,” says Mark Mathews, CTO for Gigantor.
Some solutions in the market alter portions of your AI/ML model by filtering low weighted calculations, altering the results, which may require recertification and retraining for your model to reclaim the original accuracy. In systems being used today, stock hardware solutions throw excessive memory and faster clock speeds at the problem, leading to a dead-end solution as the workload continues to increase with model complexity. GigaMACS takes your TensorFlow, Pytorch, Caffe, ONNX, or other format model and converts it for you into a highly optimized circuit that runs faster and reduces latency to under a millisecond. Now, you can run your model against 4K or even 8K video in real-time.
Need proof our concept works? Watch this short video.
About Gigantor Technologies Inc.
Don Gaspar and Mark Mathews founded Gigantor Technologies to boost neural network performance into real-time frame rates for complicated, high load Machine Learning and Artificial Intelligence models. Our first product, GigaMACS, takes your existing TensorFlow or other neural network format models and converts it to a high-performance circuit to use with existing FPGAs or custom ASICs. GigaMACS will accelerate your neural network model to keep up with your high-speed input. Have a fast full frame camera, radar, lidar? GigaMACS will make your model keep up with it without altering your results.
About Don Gaspar
The CEO, Don Gaspar, graduated with a B.S. in Physics from CSU Long Beach, an M.B.A. from the Graziadio School at Pepperdine University, and an M.S. in Distributed Systems from DePaul University. His career includes being a CTO for ISGN, Yapstone, and ImproveNet. Most notably, Don was an engineering lead who worked on the Mac OS, A/UX (Apple Unix), the Finder, and the PowerPC proof of concept at Apple Computer.
About Mark Mathews
The CTO, Mark Mathews, has developed complex and innovative IP for more than 30 years. Mark joined Microsoft in 1980 as their 17th employee and originated MultiPlan, which is better known today as Excel. Specializing in recovery of failed hardware and software projects, he also designed massively parallel machine vision circuits implementing custom and standard algorithms (e.g., LKT) as synchronal pipelines for maximum throughput at minimal power. Mark also designed and implemented proprietary tools to reduce license costs and built world-wide collaborative systems for ASIC production.