2.7 How are AI agents integrated?
7 AI agents manage $BRICS exposures 24-7
Last updated
7 AI agents manage $BRICS exposures 24-7
Last updated
ChinaAI is comprised of 7 AI agents who work autonomously to manage $BRICS exposures. While each agent is driven to act by Large Language Models (fine-tuned GPT-4 by OpenAI, Anthropic's Claude and DeepSeek's R1) they themselves employ machine learning and statistical tools to predict portfolio risk and refine CDO exposures:
Valuation Agent - Continuously calculates the intrinsic value of underlying assets in search of arbitrage opportunities. This culminates in the tracking of the tokenized CDO value over time. Examples of outputs from this agent may be found in the CDO Report. Note that the 'value' reported is a fundamentals-based (discounted cash flow) value given real world credit origination. This differs from the $BRICS price as measured in public crypto exchanges, which will include a speculative component.
Sentiment Agent - Analyses market and investor sentiment. Market sentiment is drawn from a daily assessment of 300-500 news articles. For example, the Sentiment report here aggregates 90 unique sources including Reuters, Bloomberg, Financial Times, Wall Street Journal, CNBC and local BRICS news outlets. Investor sentiment is drawn from governance votes on the protocol. The sentiment agent remains in constant contact with both the Portfolio Manager and Marketing (social media) agent.
Fundamentals Agent - Analyzes fundamental data (Financial Statement and proprietary bank transaction or credit data) and generates trading signals for portfolio exposures. Table 1 shows typical variables assessed by the agent. Explore the modus operandi in our paper, "The Impact of Bank Data in Default Prediction".
Technical Analyst - Analyzes macro technical indicators (regional, country, industry and geo-political risk for instance) and generates trading signals
Risk Manager - Calculates risk metrics and sets position limits per asset/portfolio, leveraging Copulas. For instance it estimates:
VaR (Value at Risk) at 95% confidence
Gaussian copula for normal market conditions
Student's t-copula and Levy copula for tail dependencies
Kendall's tau correlation and Inter-country sentiment dependencies to assess contagion effects.
Portfolio Manager - Makes final CDO allocation decisions, and considers investor sentiment before generating orders for trades or portfolio exposures
Marketing & Comms - Represents ChinaAI on social media and takes into account data from the Portfolio Manager. Nevertheless its key task is to keep public stakeholders socially engaged.