Myrtle.ai Transforms Financial Machine Learning with VOLLO's Unprecedented Latency Reduction

Myrtle.ai's Groundbreaking Achievement in Machine Learning Inference

Myrtle.ai, a frontrunner in the field of machine learning inference acceleration, recently announced a significant milestone for its product, VOLLO®. During the STAC® Summit in London, the company revealed that VOLLO has achieved a latency of just 2 microseconds in machine learning inference, a breakthrough recognized by the leading benchmarking authority in the finance industry.

Notable Benchmark Performance

STAC-ML (Markets) Inference sets the gold standard for solutions performing inference on real-time market data. Developed with contributions from quants and technology experts in top financial firms, this benchmark measures performance, resource efficiency, and overall quality of technology stacks capable of handling inference calculations. In this competitive landscape, VOLLO consistently outperformed all previously tested systems, making it an indispensable tool for businesses aiming to excel in the fast-paced world of financial trading.

Competitive Edge in Trading Operations

The latency achieved by VOLLO effectively provides users a competitive edge, allowing them to make quicker decisions based on more complex models. This enhancement is particularly advantageous for trading, risk analysis, and pricing activities, catering to the intricacies associated with these demanding financial operations. The ability to generate alpha rapidly places VOLLO at the forefront of a multitude of leading trading organizations.

With hundreds of thousands of hours of production trading experience, VOLLO supports a broad range of models crafted and trained using standard machine learning tool workflows before being compiled for deployment on users' hardware choices. In the testing environment, VOLLO was run on the PCIe accelerator card FBAP4@VP18-2L0S provided by Silicom, equipped with an adaptive AMD Versal™ Premium-Series SoC (System on Chip). This advanced hardware design boasts PCIe Gen5x8 support and hosts over 3.3 million programmable lookup tables (LUTs), ideal for low-latency inference applications.

Collaborative Efforts and Future Aspirations

Peter Baldwin, CEO of Myrtle.ai, emphasized the ongoing collaborative efforts with clients to further refine latencies and expand the variety and scale of models that VOLLO can handle. Together with partners including AMD and Supermicro, Myrtle.ai is keen on demonstrating how these innovations can facilitate AI-infused decisions across various market sectors.

Girish Malipeddi, Director of the Data Center FPGA Business at AMD, remarked on how VOLLO’s technology exemplifies the potential of adaptive SoCs in reimagining intelligent trading infrastructures. Similarly, Michael McNerney from Supermicro highlighted the company’s dedication to meeting the demanding workloads of the financial services industry with high-performance servers, assuring that their solutions yield optimal results for machine-learning workloads.

Moreover, Anders Poulsen from Silicom expressed pride in VOLLO’s selection of their Semiconductors for the STAC-ML benchmark tests. The Artena accelerator card demonstrates a perfect fit for VOLLO, delivering deterministic low-latency inference capabilities essential for rigorous trading workloads.

Accessibility for ML Developers

Myrtle.ai has made it easier for machine learning developers to evaluate how well their models perform on VOLLO without requiring FPGA tools or specialized knowledge. Further details are available at vollo.myrtle.ai or by contacting the company for more information on leveraging this remarkable technology.

The complete benchmark results can be accessed through the STAC report (SUT ID MRTL260323) at STAC Research.

About Myrtle.ai

Myrtle.ai is a prominent AI/ML software company specializing in providing premier inference accelerators on FPGA-based platforms from all leading FPGA vendors. With deep expertise in neural networks, Myrtle.ai has developed acceleration solutions across various industries, including fintech, wireless telecommunications, large language models, speech processing, and recommendation systems. The VOLLO product range continues to significantly push the boundaries of what's possible within the realm of financial machine learning.

In summation, Myrtle.ai stands at the forefront of technological advancement in financial machine learning by achieving unprecedented levels of inference efficiency. By continually driving improvements and collaborating with industry leaders, Myrtle.ai not only enhances operational speed but also shapes the future of intelligent trading and finance.