Arta applies machine learning techniques to certain public market quant strategies. We utilize well-established investment research methodologies from the quant fund world, and apply machine learning models for trading and allocation decisions - all with the goal of optimizing for risk-adjusted returns. Arta models a unique allocation for each investor portfolio by taking specific inputs from the investor such as investment objective, their desired risk level measured by volatility, and sector preferences. The application of ML has made it possible to execute these sophisticated strategies without needing a large team of highly-specialized analysts and traders.

Arta’s AI systems are continuously monitored under the supervision of human experts to ensure reliability, safety and control. Arta’s investment strategies are also regularly reviewed to confirm they are performing as intended. How do we ensure the AI works as designed? We rigorously test it — using out-of-sample data, simulating various market conditions, managing risk using industry-standard factor models, and finally, testing it in live markets with Arta’s own capital. Only after passing various checks is it gradually made available to Arta’s members.

Arta fine-tunes its models specifically for individual wealth management, training them on extensive datasets from sources like Barra, Bloomberg, S&P, and OneTick Polygon. These time-series datasets, commonly used by global financial institutions, contain audited financial information at both the company and macroeconomic level, ensuring Arta’s AI applications are both sophisticated and grounded in proven financial principles.