©2020 by SimpleMachines, Inc.


What is our product?

AI Computation platform that supports industry-standard frameworks like Tensorflow, ONNX, Pytorch. Also supports other analytics and ML frameworks like XGBOOST, Julia, SPARK, GATK and GraphX. World’s most efficient programmable acceleration engine.

Available as

  • PCIe card available as cloud-service

  • PCIe card available as systems for on-premise deployment

Applications and Algorithms Accelerated

Using our proprietary and breakthrough composable behavior execution paradigm, we are able to execute a diverse set of algorithms that are very high efficiency. Within deep-learning, these include convolutional neural networks, single shot multibox detectors, RNNs and LSTMs, and Transformers. In particular as AI workflows expand and include processing of various forms of unstructured data, our platform provides end-to-end acceleration when multiple of these form a single business use case. Or when a single business, uses different techniques in different workloads. Instead of peicemeal off-loaded acceleration of small components of a full-flow or requiring different acceleration techniques for each business use case. Outside deep-learning we can accelerate regressions trees, sequence alignment, and traditional data analytics done in databases.

Industries targeted and solutions we can address

Financial Services companies are currently deploying deep learning algorithms to help their businesses in the following areas.

  • Fraud Detection: real-time analytics leveraging deep learning to detect fraudulent transactions in credit card processing and loan applications, to automate and improve the KYC process, and to authenticate account access

  • Market Modeling and Analytics: natural-language processing deployed to analyze consumer and industry sentiment, forecast imbalances in the supply chain, predictive analytics on metrics like same store sales, big data quantitative analysis for stock picking and portfolio allocation models

  • Customer Service: Conversational AI to improve call center operations through the use of chatbots and virtual assistants, integrate banking functions with smart home speakers

  • Underwriting: Machine learning to make credit decisions and predict default rates

  • Back Office Operations: AI powered high frequency trading, robotic process automation to digitize paperwork and automate repetitive tasks like reconciliation and document review 

The insurance industry has been actively adopting AI to improve its operations in four key areas

  • Claims Processing: leveraging image analysis to make quick estimates and claims settlements, AI models to detect fraudulent claims, NLP to parse healthcare provider contracts to validate claims

  • Customer Service: Conversational AI to power chatbots and virtual assistants

  • Underwriting: AI models to optimize quotes and approval process, image classification to leverage satellite photos of physical properties, personalized pricing and premium discounts based on analysis of telematics data

  • Back Office Operations: Robotic process automation to streamline low-end repetitive tasks like state license verification of agents, call routing, business segmentation, customer churn analysis

Networking companies rely on AI for their two most critical product features:

  • Network Operations: AI algorithms to identify root cause performance issues, self-healing capabilities, policy enforcement, network segmentation, and monitor device health

  • Network Security: ML models for the detection of ransomware and malware attacks, threat analysis and classification, threat mitigation, and prevention of phishing and account takeovers, 

Retailers are deploying AI to advance their businesses across several different initiatives as well:

  • Product Sales: AI being used to create new in-store experiences such as smart shelves, interactive dressing rooms, and smart mirrors; ML based online recommender engines, AR/VR product displays, and image detection for powering visually-based search engines

  • Customer Service: using robotics and GPS coordinates to help customers locate products and map shopping routes, conversational AI to power online chatbots, 

  • Inventory Management: Robotics and image classification to take and scan photos of inventory, ML powered models for demand forecasting

  • Marketing: ML models to optimize ad spend and customize marketing campaigns based on online behaviors

  • Operations: ML based risk analysis, robotic process automation to power pick and pack solutions for online orders, AI models to reduce shrinkage and fraud

Healthcare is currently seeing an explosion of R&D with the goal of using AI to automate and improve a wide array of procedures and tasks.

  • Health Imaging: Trained ML models are being deployed to use image detection on a variety of health imaging tecnologies including Xrays, CAT scans, MRI’s, Ultrasounds, and Mammograms to automate basic diagnostic tasks

  • Patient Experience: Conversational AI and NLP used to help collect health backgrounds and answer basic health questions

  • Data Analytics: ML models to find patterns in numerical medical test data for basic diagnostics and patient health monitoring


Our first platform Mozart targets inference acceleration deployed through baremetal access, VM, container, or REST. Our platform for the next three generations will successively provide increasing levels of sophistication, and performance. Bach which is expected to be released in mid-2021 (currently under alpha testing) will provide distributed training across physical sites along with native quantized training. Chopin which is under development will provide direct support for AutoML. For us this means, direct Tensorflow to bare-metal hardware acceleration, eliminating the need for the many torturous manual steps today of fp32 training followed by a quantization flow, working with chip developers to optimize missing operators, fiddling around with integration of TF into other standard pipelines like SPARK etc. We are working with select industry partners on deep integration into standard software stacks.