It's taken a metric ton of work and more coffee than I care to admit to but I’ve created the start of a Machine Learning as a Service platform that gives engineers a way to train cutting edge algorithms more reliably, better and faster than ever before.
Machine learning is still in its relative infancy and creating a good algorithm can take an engineer many weeks or even months. Typically an engineer will have to make a huge array of decisions that range from deciding which features to use to the actual architecture of their algorithm. Each of these decisions will have an impact on how the algorithm will perform after an extensive training process that takes a great deal of time and computational power. Furthermore, the field can be very counter intuitive, with engineers not necessarily knowing what will work best for their algorithms until they actually try them. All of this complexity and uncertainty results in engineers being forced into a lengthy trial and error approach, that may not necessarily result in finding the most efficient solution.
Our platform encourages a different but, in my opinion, better approach to machine learning that embraces the complex nature of the field and uses it as an advantage. Instead of a slow, frustrating trial and error, the platform breaks the problem down to two distinct stages, “Pilot” and “Production”.
During the “Pilot” stage, engineers are able to quickly and cheaply train hundreds of their own custom algorithms, each using different architectures, features and hyper parameters in parallel to quickly zero in on what works best. Once the platform has helped them find the configurations that work best, they can then use our “Production” stage to train their algorithms using a larger data set, with the confidence that the final result will perform to their expectations, if not more.
The platform takes a process that currently lasts many months, even years and condenses it down to a much smaller period of time. The platform has been built with the motto “for Engineers, by Engineers” from the start and it has resulted in the development of an extremely efficient platform containing numerous innovations in both the software design and architecture.
The company is now starting an initial seed funding round seeking investment to build a world class team to bring the platform to market. While there are still a few considerable technical challenges to overcome, such as integration with GPUs and handling data sets when they approach big data limits, I’m confident a small team comprised of good engineers can find elegant solutions.
If this is of interest to you I’m both able and keen to share a business plan, present a pitch or demo at any time. If you’d just like to know a bit more, I’m also more than happy to schedule a brief call or coffee to go over what I have in more detail.
Please feel free to get in touch with us at:
or by contacting me directly.