Would you give us PolyChord to try?
PolyChord is a protected Intellectual Property, so much as we would like to, we can’t just pass the complete technology over to you casually. It is a large piece of software (a complex algorithm with its own on-board computing engine) and is custom-made to tackle very Big Data tasks. It’s also difficult for outsiders to use without some lengthy training, and it’s built in Fortran. That said, we have shared it with selected companies, with a collaboration agreement and NDA in place, but only when they have a very high level of data science competence. Most companies just don’t have the level of knowledge required to implement and deploy this powerful and effective tool. Bearing this in mind, we are willing to work with you and your team with our management and highly trained data science team, to get results. We have collaborative ways of working with you, see business models below. And we have build Python interfaces to ensure it can link to one of your own useful APIs.
What is the business model? Do you always engage with people in the same way?
We are flexible to meet your needs.
We can work for an agreed and fixed fee, established individually after looking at the task with you and defining outcomes, milestones, deliverables and timelines. Where appropriate, some organisations like the idea of creating mutual IP which can then be further exploited by both parties for commercial gain. This is in situations where the output is produced both through our unique data science, and your domain expertise in a particular area. These terms are individually agreed, but we would look to you to cover initial costs to gain an MVP and will agree on those costs after defining the task at a level which you find acceptable. In these situations, we work as close to cost price as possible and look for good profitability in the longer term in the outcoming product.
If we contribute to a project with you, who owns the I.P at the end?
We draw up a collaboration agreement to make sure both parties are in agreement and as fully protected as possible.
Do you ever do Innovate UK?
Absolutely, and we have been successful in securing funding through Innovate UK for several projects. Please contact us if you have a project on which you would like collaboration from PolyChord.
Do you get involved in Horizon Europe?
We have enjoyed being invited to Horizon Europe events (when you could still go to events) and have found some very good European collaborators there. So we will get involved if the right project is there.
Do you collaborate with RTOs?
Yes we have certainly collaborated with RTOs in the past, including Queen’s University Belfast, University of Southampton and University of Guelph in Canada, however PolyChord itself is a Limited company and thus a commercial entity, with commercial goals and although we do R&D and still enjoy a uniquely strong relationship with Cambridge University we must always look towards our commercial outcomes.
How long does it take to complete a PolyChord job?
This is highly dependent on the nature of the task at hand. Unlike other data science tools, we don’t require lengthy data preparations periods (which with some tools can take up to 18 months before even beginning the actual “data analysis”). However we go by accuracy and detail rather than speed of training. To get more of an idea of how long a project of yours might take, do please get in touch with us. We are happy to sign NDAs if you want to look at the detail.
How can I find out if PolyChord is suitable for addressing my needs?
This is dependent on the task and needs discussion to explore. We prefer to be honest, if we can’t help you we will say so. After all we are only interested in delivering results which stand-out, better than others could have delivered. To get more of an idea, do please get in touch with us. We are happy to sign NDAs if you want to look at the details in depth.
Will PolyChord work in Windows, or .NET?
We are most happy in Linux environments, that said we are working with a collaborator to get results in windows/.NET right now.
Does PolyChord work with MATLAB?
We would prefer to look at the original unfiltered data. However we are working on MATLAB integration.
What if my data varies, it’s good in some places and not in others? What about if my data is really “dirty”? Not all of my data is labelled, can PolyChord still work?
A lot of data fits this description, we are working on some tasks which fit this exactly. PolyChord has very good ability to discern between signal and noise – if you think about its origins. Naturally like any technologists we prefer data to be cleanly labelled, however we can sometimes make use of data very effectively when it is not. Like other Bayesian data science technologies, we are very good with “sparse” data sets when you don’t have very much to go on, and we are also good at “filling in the gaps” when you have missing pieces, in a way which gets results.
We’ve already tried to optimise with a different tool, and we’re not too happy with the results.
Not all optimisation tools are the same! Until now, most so-called optimisation tools have not really been up to the task. Working with people who have already attempted some form of optimisation on their data is one of our favourite situations to walk into. We find we are often able to improve upon existing work, rather than simply throwing hard work you may have already undertaken away, to get results that you’re happy with.
What about visualisation of the results?
We can often deploy visualisation tools you are already using and visualise with them with a python extension
Can you hook into my existing API?
We would need to look at it, python will often deliver a good answer.