Supercomputing, hardware systems, and software for machine learning, high frequency trading, and more, built to specification.
Modulus has built its reputation on complex, ultra-high performance solutions. We serve much of the financial industry, from JP Morgan Chase, Bank of America, and Credit Suisse to TD Ameritrade and the NASDAQ Stock Exchange. We also serve some very diverse sectors: from space exploration (NASA) to business intelligence (IBM, Shell, Siemens) to healthcare, national security, defense, and law enforcement.
From HFT trading to deep-learning A.I., our diversely skilled hardware and HPC engineers have the experience it takes to deliver high-performance solutions that harness the latest technologies.
For example, our industry-leading machine learning system is based on the IBM Power9 microprocessor, which features CAPI (Coherent Accelerator Processor Interface). The system hosts several of our FPGA-based deep learning algorithms that have been developed by Modulus using the VHDL programming language. The system hosts 1 TB of memory per socket using industry standard 64 GB DDR4 RDIMM and is capable of running complex machine learning algorithms in less than 30 nanoseconds. That's up to 250 times faster than highly optimized systems written in the C programming language. The system also features multi-port full-duplex rates of 140 Gbps at 156.25 MHz for ultra-fast optical data communication. The speed offered by this system isn't just a competitive advantage; it actually approaches the limits of physics.
Modulus offers custom ASIC solutions for applications where FPGA performance is insufficient. While ASIC solution development requires a larger up-front investment of both time and money compared with FPGA development, ASIC solutions offer up to ten times the performance of FPGA solutions.
In 2019, Modulus plans to unveil a new AI system that will be based on the Intel Nervana Neural Network Processor (NNP) L1000 ASIC chip, which will offer faster training time for deep learning models.
Supercomputers are designed to handle the most challenging simulations, analytics, and machine learning workloads. They perform enormously complex tasks: uncovering the origins of the universe, modeling trillions of neural connections of the brain, decoding patterns of protein folding that make life possible, predicting weather and financial markets, and more.
For supercomputers to be useful, they require high-performance software that's specifically developed for the architecture being used.
By providing expertly engineered software for your HPC requirements, our engineers will help you optimize your computing resources no matter what commercial supercomputing system you're running, whether it's Cray (XC and CS series), IBM (Sequoia, Mira), Fujitsu (K computer, Oakleaf-FX), Dell (Stampede), Atos/Bull (Curie, Helios), HP (Tsubame 3), Sugon/Dawning (Nebulae), or any other system.
Our engineers have extensive experience with data collection tools such as Apache Kafka and Spark; with machine learning tools including MLlib, BigDL, and TensorFlow; and with high-performance graphing tools such as the Cray Graph Engine.
Most importantly, our engineers and data scientists use well-defined, proven methodologies to approach and solve real-world high-performance computing problems.
Deep learning (also known as deep structured learning or hierarchical learning) is part of a broader family of machine learning methods. Learning can be supervised, semi-supervised or unsupervised.
Deep learning and machine learning are radically changing the way businesses operate and compete-from the way they interpret and analyze data to the speed at which they adapt to changes in the marketplace.
Companies that are heavily invested into deep learning include Man Group, Euclidean Technologies, Sentient Technologies, Twitter, AirBnB, Uber, DropBox, Google, Intel, AMD, Coca-Cola, SAP and many others.
With big data at their fingertips, companies in many sectors are using deep learning algorithms to process data the way human brains do: by transforming information into more abstract and composite representations, which enables autonomous learning.
To use deep learning to your advantage and change your trajectory, you need a solution built and backed by engineers with expertise in machine learning, data science, and HPC.
At Modulus, we've been immersed in machine learning and high performance computing for more than 20 years. In fact, we designed many of the deep learning technologies in use today by industry and research pioneers. We understand what deep learning can do, how it's evolving, and how to put it to work in practical, high-value ways to solve problems.
Modern technology, high-performance computing, machine learning, and innovation: these have been our competitive strengths for over two decades. This is why leading organizations around the world rely on Modulus.
By partnering with us, you'll get the outcomes you want under a success-based fee model that mitigates risk.
Next-Gen HPC Application Development
For over 20 years, Modulus has been developing and maintaining the largest and most valuable source code repository on the planet for financial applications, machine learning, and high-performance computing. Using our innovative and reusable solution accelerators and frameworks, our world-class team will work with you to create unrivaled next-gen applications.
HPC application development requires expertise in parallel computing and extensive knowledge of hardware systems. When software is too slow for an application, our hardware engineers lend a hand.
High Performance AI Software Development
Machine learning is a profound technology platform shift that's helping to shape our future. For predictive analytics, planning and optimization, perception and situational awareness for trading, sales predictions, marketing and advertising optimization, manufacturing, and more, Modulus has real-world experience in applying A.I. and machine learning to business problems across many verticals with exceptional results.
Our expertise includes real-time A.I., deep learning using neural networks, genetic algorithms, and bio-inspired forms of A.I. Using these technologies, we've developed advanced trading systems, UAV autonomous flight controllers, self-driving automobile systems, healthcare diagnostics, security and defense systems, social mediate sentiment systems, and much more.
The Modulus Social Media Sentiment Analysis System uses deep learning neural networks to extract mood and emotions from millions of social media posts every day. Running uninterrupted for nearly a decade, the system has created multiple petabytes of historic data. Clients include top hedge funds and financial institutions.
HPC Trading Systems
With the increase in low-cost computing power, many trading systems use multidimensional nonlinear modeling such as deep learning with neural networks, kernel regression, nonparametric modeling, dynamic programming by means of genetic algorithm, rough set theory, and other processor intensive algorithms.
Our engineers bring decades of experience and the firsthand knowledge needed to select the right tools for creating profitable, dynamic trading systems for equities, futures, options, bonds, cryptocurrencies, and other tradable assets.
Under strict confidentiality, our engineers have discretely developed numerous trading systems for some of the largest and most successful hedge funds in the industry. We have the first-hand knowledge required to select the right tools for creating profitable, dynamic trading systems that can support HFT.
Our Buy Side solutions consist of high frequency data acquisition and management, algorithmic trading system development for equities, futures, options, bonds, crypto, OTC and other tradable assets, dynamic position sizing algorithms and risk management systems, order routing including dynamic smart orders, big data analytics and mining including analysis of unstructured data, and execution Management System (EMS) design and development.
Sell Side solutions consist of advanced market making and hedging systems, advanced price construction algorithms, smart order routing for optimal execution to split orders between multiple venues, auto trading system strategy design using A.I. for high frequency, statistical arbitrage, inter-market and multi-time-frame trading, high throughput network design and implementation for HFT with latency as low as 20 nanoseconds.
HFT, DMA, Quant Systems