<aside> đź’ˇ How to be up and running with the backend of a world-class quant fund in minutes.

</aside>

Introduction

Judge Research’s SDK was built by P3k, the ecosystem’s first participant organization. They built the SDK in order to have a system that was:

  1. Linked up to an extra-reliable, real time & tick-by-tick data for spot, futures and options markets. That way, they could start feature engineering effortlessly. For example they could start a new script with something like x1 = min1.binance.BTC.spot.close instead of thinking about cumbersome API calls, data cleaning, and so on.
  2. Pre-loaded with all of the machine learning and graphing Python packages you might expect, so research prior to submitting data was easy and convenient.
  3. Separate from their main backend, so the regular submission of features to the decentralized fund would not be interrupted if their backend went down.
  4. Able to translate data in a matrix into the properly-formatted JSON that Judge Research’s API expects with a single shift+enter. Likewise, it needed to handle the initial submission of many thousands of observations and the subsequent submission of live data with a single click.

They have open-sourced their build, so your fund can start participating in the Judge Research ecosystem in minutes.

What Is The SDK?

The SDK is a group of functions in a Jupyter Notebook that comes pre-loaded with most of the packages you would like. These functions make it easy to call high-quality market data, provides a sandbox to do your feature engineering & related research, and then automates the process of sending in the features you engineer.

Importantly, it also makes it easy to build your research code base in dynamic documents into which you can plug live findings from Judge Research as iframes. This might be considered more advanced usage, so it likely is better to start with the basic workflow.

Video Tutorial

If you prefer to watch a two minute video on building the SDK, check out the below. At the end of the video, you’ll find another tutorial on how to submit features.

https://www.youtube.com/watch?v=-R9FTw-mNZg&list=PL1dveJ4wpejpgHABLflKA-_qBp5fT0Usl&index=6

Instructions

It is easy to run the SDK locally, on a serverless notebook service, or on a cloud-based machine learning instance.

Of those options, only the latter is likely reliable enough for the many-times-a-day, months-on-end feature submissions you’ll need to participate in the decentralized systematic fund.

1. (Optional) Play Around: Launch the SDK on a Serverless Notebook With One Click