Artificial intelligence is gradually becoming part of the financial markets infrastructure. However, Invorium proposes viewing AI not simply as a tool for analysis or forecasting, but as the foundation of a specialized computing system capable of independently processing market data and executing algorithmic operations.
This concept forms the basis of IVA (Invorium Virtual Assistant) — the company’s new product, presented by CMO Marc Ferrer.
Unlike common AI solutions designed for user interaction, IVA was developed to operate directly within cryptocurrency markets.
From Analysis to Execution
The primary task of IVA is to process large volumes of market information. The system analyzes prices across different trading platforms, compares liquidity, monitors trading volumes, and identifies price discrepancies.
Arbitrage plays a particularly important role. If the same digital asset is traded at different prices across multiple platforms, the algorithm evaluates the possibility of executing a transaction. In doing so, it takes into account fees, available capital, potential slippage, trade volume, and other constraints.
This approach involves not merely identifying a price difference, but conducting a comprehensive assessment of the economic feasibility of a transaction.
Specialized Hardware Platform
IVA exists not only as software. The physical device represents a computing node designed for AI analytics and automated operations involving digital assets.
The declared architecture is based on Neural Matrix VII, featuring 144 energy-efficient RISC cores and four 128-bit vector SIMD blocks per core.
The technical configuration also specifies 384 GB of Photonic NVM, internal memory bandwidth of up to 16 TB/s, and interprocessor connectivity of up to 900 GB/s.
For AI computing, the system offers performance of up to 36 PFLOPS in FP4 and up to 18 PFLOPS in FP8.
Operating Environment and Risk Control
The software foundation of the device is IVA OS. It includes Neuro Kernel, Memory Fabric, Quantum Scheduler, Execution Engine, Risk Matrix, and Cluster Sync.
Each component is responsible for a specific part of the computing process — from task distribution to node synchronization and the control of trading constraints.
According to Marc Ferrer, the fundamental principle behind the project is not the maximum number of operations, but a systematic approach to identifying economically justified trading opportunities.
At Invorium, IVA is viewed as a technological platform whose architecture is designed for further development and operation across various trading strategies and markets.
