Myrtle.ai Achieves Unprecedented Latency Reduction in Financial ML Using VOLLO

Myrtle.ai Sets New Benchmark for Machine Learning in Finance

In the ever-evolving landscape of financial technology, myrtle.ai stands as a beacon of innovation with its groundbreaking VOLLO product. The recent announcement from the company highlights a significant milestone: a reduction in latency for financial machine learning (ML) inference benchmarks. This comes as a game-changer for trading firms looking to gain an edge in a competitive market.

The STAC Benchmark Unveiled

During the STAC Summit held in London, myrtle.ai revealed that its VOLLO product, a solution utilizing Field Programmable Gate Array (FPGA) technology, has been audited by STAC, a premier benchmark authority for the finance sector. The significance of this benchmark cannot be overstated, as it is tailored specifically for real-time market data inference, designed by quants and technologists from leading financial institutions.

VOLLO has demonstrated an impressive ability to achieve latencies as low as 2 microseconds in the 99th percentile across various benchmark tests. This marks a remarkable leap over earlier systems, effectively halving the previous latency records. The implications of such a swift response time are profound: users can now execute more sophisticated models and make informed decisions at an unprecedented pace.

Performance Meets Efficiency

The VOLLO technology has been validated through extensive use, with hundreds of thousands of hours of production-level trading. It is already generating alpha for many top-tier trading firms, which have integrated it into their existing machine learning workflows.

The technological prowess of VOLLO is displayed in the system configuration used for testing, which included the FBAP4@VP18-2L0S PCIe accelerator card from Silicom, equipped with an AMD Versal Premium series VP1802 Adaptive SoC. This combination not only meets the demands of low-latency applications but also boasts advanced features like PCIe Gen5x8 capabilities and over 3.3 million programmable lookup tables (LUTs).

Peter Baldwin, CEO of myrtle.ai, expressed his excitement about working alongside AMD, Silicom, and Supermicro, emphasizing the collaborative effort that allowed them to showcase how their technologies could transform quantitative trading landscapes.

The Future of Financial Markets

As financial markets increasingly rely on rapid data processing and real-time decision-making, the prevalence of AI systems capable of interpreting vast datasets and reacting swiftly will become vital. Girish Malipeddi, from AMD, articulated this future vision, stating that the integration of advanced AI technologies will be instrumental in shaping market infrastructure.

Further amplifying this sentiment, Michael McNerney from Supermicro noted the importance of optimized server performance for demanding workloads in the finance domain, illustrating a broad commitment across industry players to innovate.

Silicom's involvement highlights the importance of choosing the right hardware for meeting stringent latency demands, showcasing how tailor-made solutions can yield high performance under demanding financial workloads.

Accessibility for Developers

Myrtle.ai is also keen on inclusivity in ML development by providing access to their VOLLO platform for developers. They can evaluate model performance without needing specialized FPGA tools or technical expertise, simplifying the process and fostering innovation.

To explore VOLLO further, interested individuals and companies can obtain more information on its functionalities and how it can enhance their trading performance at vollo.myrtle.ai.

Overall, the advancements introduced by myrtle.ai with VOLLO cannot only be seen as a technological triumph but as a strategic enabler for firms aiming to leverage AI for profound financial insights and operations. The journey towards ultra-low latency AI inference has only just begun, and myrtle.ai is at the forefront of leading this charge.