Custom machine learning research platform

[fan27]

Later this month I will have completed my machine learning research platform proof of concept. It is written in Golang and supports multiple time-frames, is easily extended and super fast. It is designed to use different feature combinations to find trade-able patterns. I have developed techniques to reduce a combination permutation set from 80,000 to the combinations I am really interested in to 2,000 in one example. Each iteration in my current code takes about half a second (on lap top) and that is with little optimizations. So that is about 16 minutes for a complete run. Also, the platform is configuration driven where JSON data can drive behavior lending itself to easily add a slick web front end.

The point of this thread is to seek advice on how I could get funding to complete this project. Of course I could keep chugging along developing this in my spare time, but I would love to have a year to just focus on this project with the ultimate goal of starting a hedge fund or joining an existing hedge fund that was interested in the technology.

Assuming I can create a very compelling presentation on how this platform will be able to offer a competitive advantage over the typical tools used by most hedge funds, who would be the audience for this presentation?

Also, I could use a portion of my trading capital to fund the development with me leaving my job. I just need some way to gauge interest in the project to justify that move.

thanks
fan27

 

[IAS_LLC]

You need to provide more details if you wish to get useful feedback.

What EXACTLY does it do? What value will it offer to your investors? How will they recover their investment? If your answer to that question is clear, believable, and compelling... than it should be pretty easy to attract investors. Saying you're using "machine learning" to perform dimensionality reduction isn't compelling.... it is trivial with open source tools.

 

[fan27]

What value will it offer to your investors? If your answer to that question is clear, believable, and compelling... than it should be pretty easy to attract investors. Saying you're using "machine learning" to perform dimensionality reduction isn't compelling.... it is trivial with open source tools.

I totally agree. The presentation will obviously need to differentiate the technology from currently available alternatives. In a situation like this, would a hedge fund already need to be established to attract investors?

 

[IAS_LLC]

I totally agree. The presentation will obviously need to differentiate the technology from currently available alternatives. In a situation like this, would a hedge fund already need to be established to attract investors?

I'd say yes. I'm not going to give money to "some guy" without a track record. Especially if my money is going to be used to develop a system, not trade an already established system.

Why do you need other people's money ? Imo, you're better off trading your own capital to establish a track record . If development time reduction is your motivation....outsource pieces.

 

[fan27]

That makes sense. I would be more than willing to pay to outsource portions of the development. The challenge would be finding someone that could actually do the work without me having to spec everything out everything which would likely take as much time as me doing the work myself.

 

[wintergasp]

I'd say yes. I'm not going to give money to "some guy" without a track record. Especially if my money is going to be used to develop a system, not trade an already established system.

Why do you need other people's money ? Imo, you're better off trading your own capital to establish a track record . If development time reduction is your motivation....outsource pieces.

I believe that fan27 is trying to sell or lease a trading platform that you can use to develop your own systems, such as Deltix, MultiCharts, Quantconnect etc. A lot of professional managers have an interest in using such platforms to develop tradable strategies and won't ask for any trading track-record as what really matters is the technical features offered.

 

[IAS_LLC]

I believe that fan27 is trying to sell or lease a trading platform that you can use to develop your own systems, such as Deltix, MultiCharts, Quantconnect etc. A lot of professional managers have an interest in using such platforms to develop tradable strategies and won't ask for any trading track-record as what really matters is the technical features offered.

He specifically stated his desire is to start a fund. I agree with your premise, if that was indeed his intent, however

 

[fan27]

@IAS_LLC is correct. Ideally I would like to start a hedge fund around the technology I am developing. Getting the software to a polished state that could be sold to end users would dramatically increase development costs. That being said, I am open to that idea.

 

[ofthomas]

That makes sense. I would be more than willing to pay to outsource portions of the development. The challenge would be finding someone that could actually do the work without me having to spec everything out everything which would likely take as much time as me doing the work myself.

unless you spec everything out, and provide clear definitions around what you want done, outsourcing wouldn't work for you... how can any developer create something that you want, unless you tell them what is it that you are after and what the output should be, etc? little definition leads to scope issues and cost over-runs... its rather simple.

 

[fan27]

unless you spec everything out, and provide clear definitions around what you want done, outsourcing wouldn't work for you... how can any developer create something that you want, unless you tell them what is it that you are after and what the output should be, etc? little definition leads to scope issues and cost over-runs... its rather simple.

Yep...my thoughts my exactly.