Longevity

The automation pipeline consists right now of 2 loops.

One outer, which has to run first, then one inner.

Production at this moment is junior level, the program was set up in >7 days.

Goal is to deploy this at any firm, essentially you could hire me as a junior qfin researcher, and if it is okay I would use this automation as output.

The project is in better use though if a team works on it, when outer loop enables the inner loop to produce some satisfying results, that's where the magic looks to be.

Would want to post a CMD run here but it would reveal the structure, so if a firm wants to see this, alright under NDA.

Since this produces junior level research and can run 24/7, and was set up relatively quickly, the main benefit to firms would be less need for junior qfin researchers, of course keep a few since this automation is AI and humans need to manage it.

= Lower.Cost.For.Firm per year in research

= Hopefully.Better.Output than human level research (the inner loop is the main key)

&basically end game is finding out if completely new systems from inner loop is best ROI or if small improvements to already existing systems is better long term.

If new systems -> automate that too, make it learn what parts mined in the inner loop go best together, sort of like Lego.

The research is high level, sometimes I read research in fields I don't understand or know, out of boredom & curiosity, I like the structure of phase 1-4 when conducting & that's the main idea with the loop(s).

I am 25, previous experience in accounting, banking & in my free time I've studied programming, machine learning & human behavior.

My deepest interests are diving into projects head first, and figuring out things as I go.

Current focus is "deep work", getting the most work out of a time given, and productivity.

In trading I use discretionary systems at current moment, looking into automation, have built some programs that make the systems easier to use. 

Structure or focus right now is generating potential  α  signals.

Successful. - generates >10 per run (5-10min per run)

Runs and gives potential  α 's, but the backtesting is lightly hypothetical.

Keep in mind this is done on my infrastructure, on a decent pc but still.

Improvement area would be extra validation or testing the  α 's live.

&no I did not ask GPT to generate 10 potential α, I have built this automation in 3-4 different forms, each open to improve in different ways, focusing on α  is most interesting in my opinion.

Below is an example of hypothetical α generated in a coffee break.