Realtime machine learning

Over the past few years, we have been developing a CAI-Server (Customized AI Server Software as a Service). The CAI-Server is a unique multi-component software platform that gathers streaming data, make sense out of the data to produce results that allow customers to make data-driven decisions.

On top of the CAI-Server, EZNumeric has been developing a suite of software that can be used by crypto startups. This way crypto startups do not have to start development from scratch but can make use of our turn-key infrastructure, thereby lowering risks and reducing development time by at least 2 years.

Read More

Over the past few years, we have been developing a CAI-Server (Customized AI Server Software as a Service). The CAI-Server is a unique multi-component software platform that gathers streaming data, make sense out of the data to produce results that allow customers to make data-driven decisions.

On top of the CAI-Server, EZNumeric has been developing a suite of software that can be used by crypto startups. This way crypto startups do not have to start development from scratch but can make use of our turn-key infrastructure, thereby lowering risks and reducing development time by at least 2 years.

Motivation

In the crypto space, 2018 was the year of the ICOs, Initial Coin Offerings. Everywhere you looked, crypto and blockchain companies were launching new tokens. More than one thousand ICOs have raised about $20 bln. However, the ICO market has slowed slightly in 2019. There are several reasons for the slowdown, a.o. lower cryptocurrency prices, fear for a scam, more careful investment. Moreover, there is a new crypto funding raising method: a Security Token Offering or STO. PWC & CVA expect STO volumes to grow in 2019 and 2020, as the adaption of the funding method will increase. 

Both ICO and STO face the same issues as a non-crypto startup: how to build a team, how to build a company, how to build a product. According to a well-known statistics: 90% of startups fail. One of the main reasons is inadequate product market (more than 50%). 

Problem/Solution

Most startups in crypto/blockchain space are at the idea stage. This brings a lot of risks for investors as well as for startups. 

We offer a Software as a Service (SaaS) product called CAI-serverCustomized AI server –  that allows shortened development time for a startup at least by 2 years, therefore minimizing risks for investors and allow startups to focus on their idea built on a solid fundament – our product CAI-Server. 

CAI-Server

We offer a SaaS (Software as a Service) Customized AI Server – CAI-Server.

In a nutshell, the CAI-Server gathers streaming data, make sense out of the data (analysis, forecasting) and produces results that allow customers to make data-driven decisions. Built around Kafka, our software is fault-tolerant and scalable. Kafka also makes it possible to process data across data centers.

Analysis server
CAI-server

CAI-Server is a unique innovative software platform consisting of the following components:

  • Proprietary web-servers responsible for a secure connection and data transfer
  • AI/Machine Learning library responsible for data analysis and forecasting, consisting of known and proprietary innovative machine learning techniques
  • Multiple databases to store encrypted data and intermediate results
  • Connection to external data sources from the web (www) such as social networks for sentiment analysis
  • Connection to data sources (such as exchanges) and customers via Kafka (widely used distributed streaming platform) or REST API
  • Connection to multiple blockchains (Bitcoin, Litecoin, etc..)

Features:

  • Stream and process data in real-time
  • Written entirely in C++ for efficiency
  • Embedded multithreaded secure webserver
  • Embedded multithreaded secure REST API
  • Embedded multithreaded secure WebSocket server
  • Documentation is a high-quality PDF document that is automatically generated (and emailed) on demand
  • Supports simultaneous database instances of multiple types
  • Automatically inform the user by email in case of a software crash (heartbeat mechanism)
  • Embedded dashboard web-page for administrators
  • Broadcast feature to send commands to all backends 

Crypto Currency Application

What does it mean specifically for a cryptocurrency application?

In the crypto space our product CAI-Server allows to 

  • Collect the cryptocurrency prices from several exchanges, 
  • Calculate widely used and proprietary market indicators 
  • Forecast value of a cryptocurrency 
  • Connect to a blockchain
  • Perform analysis on the blockchain (e.g. security analysis, data fraud)
  • Automatically alert users based on predefined events

Added Value

  • Shorten development time for startups 
  • Reduced risk for investors of startups
  • Enable data-driven decisions
  • Enable scalability 
  • Seamlessly connect to our CAI-Server and make sense out of your data

Interested?

Contact us for more information info@eznumeric.com, we will make sure to come back to you within 24 hours.

Big Data Solutions & Machine Learning

Did it happen to you that when someone mentions big data, you think ”I want to do it too! But I need to get a cluster, maybe even a Hadoop cluster, I need to hire a Data scientist but wait a year before getting results”. That sounds like a big investment and you are not even sure it will pay out. You might not know where to start.

EZNumeric can help you to setup and execute a big data trial/proof of concept/feasibility study to solve your problem using your data.

Read More

Has it ever happen to you that someone mentions big data, you think: ”I want to do it too! But I need to get a cluster, maybe even a Hadoop cluster, I need to hire a Data scientist but wait a year before getting results”. That sounds like a big investment and you are not even sure it will pay out. You might not know where to start. Do I boost my marketing with big data technology, do I want to know about my customers? Do I want to automate a process? Choose the optimal set of parameters? Base business decisions on your data. Wanting to make use of the big data in your business and don’t know where to start?

EZNumeric can help you to setup and execute a big data trial/proof of concept/feasibility study to solve your problem using your data. Using a 5 steps approach of the big data framework, Big Data will be demystified for your business:


Data points from 2 classes and the decision boundary made by support vector machines (SVM) algorithm
  1. Identify area of improvement, a business need, a question
  2. Assess needed input data
  3. Find a storage solution of the data
  4. Chose appropriate analytics software
  5. Visualize/report the outcome

Areas of expertise

  • Machine Learning
    • Neural networks, Deep Learning
    • Regression (linear, logistic)
    • Support vector machines (SVM)
  • Dimensionality reduction / model order reduction
  • Distributed learning: in-house Parallel Neural Network

Programming languages

  • Python
  • R
  • Matlab/Octave
  • C/C++

 

Optimization & High Performance Computing

Are you not satisfied with the performance of your modeling or simulation software? Wanting to accelerate your computations? Wondering which parallelization technique and parallel architecture to use?

EZNumeric can advise you on how to parallelize your algorithm/software and what hardware setup and parallelization techniques will be the most beneficial to achieve a significant speedup.

Read More

Are you not satisfied with the performance of your modeling or simulation software? Wanting to accelerate your computations? Wondering which parallelization technique and parallel architecture to use?

EZNumeric can advise you on how to parallelize your algorithm/software and what hardware setup and parallelization techniques will be the most beneficial to achieve a significant speedup.

Areas of expertise

  • Parallelization techniques using
    • OpenMP
    • MPI
    • GPU(s)
  • Scientific software optimization
  • GPGPU in
  • Heterogeneous Map-Reduce: in-house parallel framework

Programming tools

  • OpenMP (many-cores, XeonPhi)
  • MPI
  • GPGPU : CUDA, OpenCL
  • Socket programming

Publications

| Performance and scalability of finite-difference and finite-element wave-propagation modeling on Intel’s Xeon Phi
| A comparison of continuous mass-lumped finite elements with finite differences for 3-D wave propagation
Reduction of computing time for least-squares migration based on the Helmholtz equation by graphics processing units
Closing the performance gap between an iterative frequency-domain solver and an explicit time-domain scheme for 3D migration on parallel architectures
3D Helmholtz Krylov Solver Preconditioned by a Shifted Laplace Multigrid Method on Multi-GPUs
GPU implementation of a Helmholtz Krylov solver preconditioned by a shifted Laplace multigrid method

If you want to have a copy of our publications, please send us an email or fill in the contact form.

Numerical Modelling & Simulations

Are you looking for ways how to model a physical process or simulate your data? Do you want to solve a differential equation or a system of those? Are you trying to speed up your algorithm?

At EZNumeric we have 15+ years of experience of developing numerical models and simulations as well as finding an efficient solution.

Read More

Are you looking for ways how to model a physical process or simulate your data? Do you want to solve a differential equation or a system of those? Are you trying to speed up your algorithm?

At EZNumeric we have 15+ years of experience of developing numerical models and simulations as well as finding an efficient solution.

Areas of expertise

  • Numerical methods to solve large systems of linear equations, including preconditioner techniques (e.g. Krylov methods, multigrid)
  • Optimization / problem minimization algorithms (least squares, inversion)
  • Elliptic, parabolic and hyperbolic differential equations (wave equation, Helmholtz, streamline reservoir simulation, geoelectric modelling)
  • Compression: lossy or lossless
  • Finite differences, finite elements (continuous and discontinuous), finite volumes, meshless methods (e.g. material point method)

Programming languages

  • C/C++ object oriented
  • Fortran
  • Python
  • Matlab/Octave

Section through the tetrahedral mesh for finite element computations of the wave propagation. Made in collaboration with W.A. Mulder, S. Minisini, A. Kononov


Coarse grid correction coefficients for multigrid method in 3D
Fourier transform of frequency domain wave simulation with Marmousi velocity model

The following pictures shows the wavefront of a moving sound wave in a room. The blue surfaces (cube) and back-wall absorb the sound wave.

The animation can be found here. See the gallery for more example of computer generated images.

Publications

| A comparison of continuous mass-lumped finite elements with finite differences for 3-D wave propagation
| Time-stepping stability of continuous and discontinuous finite-element methods for 3-D wave propagation
| Local time stepping with the discontinuous Galerkin method for wave propagation in 3D heterogeneous media
| Stability and performance of the SIPG and IIPG finite-element methods for wave propagation
| Solving the 3D Acoustic Wave Equation with Higher-order Mass-lumped Tetrahedral Finite Elements
| A multigrid method with matrix-dependent transfer operators for 3D diffusion problems with jump coefficients
3D Helmholtz Krylov Solver Preconditioned by a Shifted Laplace Multigrid Method on Multi-GPUs
Reduction of computing time for least-squares migration based on the Helmholtz equation by graphics processing units

If you want to get a copy of our publications, please send us an email or fill in the contact form.