#Cambridge #United Kingdom #Myrtle.ai #VOLLO #Napatech

Myrtle.ai Unveils VOLLO for Microsecond ML Inference

In a groundbreaking announcement, Myrtle.ai, a renowned leader in machine learning acceleration, has unveiled its latest offering: the VOLLO® inference accelerator, now optimized for deployment on the Napatech NT400D1x series of SmartNICs. This collaboration marks a significant advancement in the field of machine learning, particularly for applications that demand ultra-low latency. With VOLLO, users can achieve record-breaking inference latencies, often under one microsecond, a feat not commonly attainable in the industry.

The Need for Speed in Machine Learning

The demand for rapid machine learning (ML) inference has surged across diverse sectors, especially in areas such as financial trading, cybersecurity, and telecommunications. In these contexts, every millisecond counts, and thus, the ability to perform inferences at unprecedented speeds can lead to substantial benefits in efficiency, security, and profitability.

Myrtle.ai's VOLLO has been engineered to meet the challenges posed by current market needs. By processing ML models right at the network edge with a SmartNIC, VOLLO minimizes latency significantly, which is paramount for applications requiring real-time data analysis and decision-making capabilities. Popular ML frameworks, including LSTM, CNN, MLP, as well as Random Forest and Gradient Boosting Decision Trees, can be seamlessly executed on this platform.

How VOLLO Works

VOLLO's sophisticated architecture leverages the inherent capabilities of Napatech's SmartNICs, enabling it to process data streams with lightning speed. This optimization allows businesses, particularly in the financial sector, to leverage machine learning for automated trading strategies that react instantaneously to market changes.

Peter Baldwin, CEO of Myrtle.ai, expressed his enthusiasm regarding this collaboration, stating, “We are thrilled to partner with a global leader in SmartNIC distribution to provide unmatched low-latency performance for ML inference. Our clients are actively seeking to push the boundaries of latency reduction, and this new version of VOLLO is designed to help them realize that vision.”

Enhancing Business Capabilities

The introduction of VOLLO not only aims to enhance performance but also seeks to broaden the operational capabilities of various industries. As Jarrod J.S. Siket, Chief Product Marketing Officer at Napatech, indicated, “We understand that leading latencies in STAC® ML benchmarks can genuinely enhance our customers' offerings in the financial market, especially as they increasingly adopt ML for automated trading.” He further noted that the VOLLO compiler is designed to simplify the usage of their SmartNICs for machine learning developers, thereby enriching their product and service portfolio.

Myrtle.ai has made the ML-oriented VOLLO compiler available for download at vollo.myrtle.ai, where interested parties can explore the specific latencies achievable with their models on Napatech's NT400D1x series SmartNICs.

About Myrtle.ai

Myrtle.ai is a prominent AI and machine learning software company, recognized for delivering state-of-the-art inference accelerators across FPGA-based platforms from leading FPGA providers. With expertise in neural networks spanning the entire range of machine learning models, Myrtle.ai has created accelerators tailored for fintech, natural language processing, and recommendation systems, among other applications. The impact of their innovations continues to resonate across industries looking to enhance efficiency and responsiveness in their operations.

In conclusion, the launch of VOLLO encapsulates a significant leap forward in machine learning capabilities, promising to redefine the standards of speed and efficiency in inference tasks. As industries increasingly embrace AI solutions, Myrtle.ai and Napatech are poised at the forefront of this transformative wave, ready to support businesses in maximizing their technological investments.