Arta applies machine learning techniques to well-established investment research methodologies from the quant fund world. Machine learning is also used for trading and allocation decisions with the goal of optimizing for risk-adjusted returns. Investors provide inputs such as desired risk level, and sector preferences. Arta uses this input to model the appropriate allocation for each investor.

Arta combines specialized AI techniques depending on the task. While Large Language Models and agentic techniques are great at reasoning, research and communication, other machine learning methodologies play an important role in investing and financial applications. For example, the deterministic nature of convex optimization is more appropriate for portfolio construction strategies that dynamically evolve over time.

Arta’s AI systems are consistently monitored under the supervision of human experts to drive reliability, safety and control. Arta’s investment strategies are also regularly reviewed to confirm they are performing as intended. We rigorously test Arta AI using various techniques, including out-of-sample data, simulating various market conditions, and managing risk using industry-standard factor models. Finally, certain investments are tested in live markets with Arta’s own capital. Only after passing various checks are they gradually made available to Arta’s members.