With all the excitement around AI, it has yet to make a significant impact to date in the crypto marketplace.  This is undoubtedly about to change as its combination with blockchain technology seems to be a mutually beneficial partnership.


Blockchains possess the capability of scaling AI since it’s able to secure storage and share paths of data, whereas AI generates insights from this data in order to create further value.  Moreover, AI’s ability to analyze large amounts of data and machine learning capacity fits well into the future developments of blockchain technology.


Technically speaking, the world of AI (and even blockchain to some extent) is truly complex and abstract for the uninitiated.  In order to see the bigger picture, visualize a kind of interwoven tapestry composed of various interconnected agents, which through their collaboration form a type of self-sustaining model.  Furthermore, these models get stacked on each other as layers upon layers begin accumulating and a comprehensive, algorithmic, and artificially intelligent system arises.


As you could imagine, building out this infrastructure requires an incredible amount of research and work on the back-end.  However, with the incredible foundation it provides, it helps to propel forward the integration of blockchain and AI together.


Below are a few examples of the architectures we have developed and are continuing to build upon.


Backtesting and Simulation Framework

Backtesting (testing a predictive model on historical data) and Simulation (implementation of those models) are important tools used in order to test the viability of trading strategies.  


Simulation framework is intended to stimulate multi-agent trading and market making activity within a configured environment.  They act as either liquidity providers posting bid and ask offers on a limit order book or as liquidity takers executing market orders against the book.  


Backtesting framework is similar to the simulation one though prices are taken from either real-time or historical data.


AI Oracles

“On-chain” data may be used for applications based on AI.  As an important part of automated, algorithmic trading deals with price predictions, then it stands to reason that it would be important as well for the algorithms built in AI.  Indeed, price predictors act as a key component of the overall suite of AI oracles.  Below are some of the examples 


Portfolio Planner Oracle: uses accumulated on-chain market data and gets predictions from the price predictor oracle in order to suggest the assets to be included in a long-term investment portfolio given an investment goals, terms and target level of risk.


Strategy Evaluator Oracle: uses on-chain data and price predictions to evaluate different strategies and parameters in order to select optimal set of trading, market making or holding strategies better suited to maintain a portfolio given current market conditions. Back-testing frameworks are also incorporated here.


Pool Weighting Oracle: uses data and predictions to weigh on market instruments and adjust portfolios in order to minimize the mid-term risks such as impermanent loss.


Signal Generator Oracle: takes predictions of current prices and sentiments to generate signals for liquidity provision, trading applications, and smart contracts.


Sentiment Watcher Oracle: monitors news feeds on social media to gauge overall crypto buzz, providing inputs for both price prediction and signal generation.


Smartpools and Portfolio Optimization with AI

Active portfolio management based on AI can be done off-chain with a Binance limit order book or on-chain with smart contracts like Uniswap and Balancer.  Price predictions and sentiment analysis also factor into the optimization process.  This model will be integrated into Smartpools on Smartdex.


Maker 2.0 and Adaptive Multi-Strategy Market Making Agent


Market making is somewhat more complicated because you have to account for not only time, volume, and activity but also for spread and fees.  Due to this, additional information must be accumulated first in the Operational Space Model for Market Making (as seen in the image below) before it can be funneled into the market making agent.  Maker 2.0 will be incorporating both of these architectures.


Have fun and happy trading with Autonio!