IVI AI

IVI AI deploys capital across liquid futures, global equity indices, and G10 foreign exchange markets. The firm avoids illiquid private markets and long-lock-up structures, focusing exclusively on exchange-traded instruments where its machine-learning models can ingest high-frequency tick data. Commodity exposures are concentrated in energy and metals futures where supply-chain data signals provide an informational edge.

The firm builds end-to-end neural network architectures rather than imposing human-derived linear factor constraints on models. Researchers rotate across signal domains quarterly to prevent overfitting on any single dataset, and portfolio construction proceeds without discretionary manager override. This contrasts with many quant platforms where PMs manually blend signals or veto model-driven allocations based on macro views.

Three data verticals feed the firm's models: tick-level exchange data from over 40 futures markets, alternative datasets including maritime transponder feeds and consumer transaction panels, and unstructured text parsed from central bank statements and earnings call transcripts via fine-tuned large language models. The NLP pipeline focuses on policy sentiment extraction from Federal Reserve and ECB communications.

The firm targets mid-frequency holding periods that sit between high-frequency market-making and fully discretionary macro — a zone deliberately chosen to allow capacity growth without immediate alpha erosion. Human oversight centers on capacity monitoring and tail-risk controls but does not override model-driven trade execution.

The firm's capital structure and investor base are not publicly disclosed. Its operational model resembles systematic hedge funds that train models on proprietary data infrastructure, though the absence of publicly reported regulatory filings makes the external-versus-proprietary capital split unverifiable from open sources.