rixel
About us Projects Hire us
DOF
DoF is a highly scalable dataset format which helps deep learning scientists to work with foreign and/or sensitive data. DoF provides fast dataset sharing and data-secure at the same time. Users can save datasets in DoF with additional information like credits and legal notices. Since DoF is scalable, it provides the possibility to store preprocessed images or trained neural network models as well. In this case, a complete AI workflow can be placed in one DoF file with appropriate credits and licenses.
Medical Gateway Platform is a data-focused solution to connect different medical devices and platforms with real-time secure and private data sharing. The goal of this project is to provide a unified data and software environment for different devices and platforms. With this environment users can collect data or control the devices easily. A good unified solution integrates ventilator systems and heartbeat sensors at the same time.
MGP
aaion
Aaion is a human circadian management system to increase the user’s performance and ensure better daily comfort. It covers different levels of application: user and server levels. Shift is not only a classic work shift system. Server level’s administrators are able to create different day periods based on templates and monitorize the states of connected users’. Day periods can be different than 24 hours, so it fits for the needs of a long time space travel. System is based on the human circadian cycle. Our solution tries to fit the daily activities or time zone changes to the users’ circadian rhythm as best as possible.
HealthData Labelling System is a software as a service solution that helps digital health or medical service companies to integrate new data from smart devices without significant investment. It collects, cleans and labels data from smart- or health devices. A lot of raw data comes from different smart devices with different tagging and labeling methods. Some sensor data arrives without any additional label, while some services (like fitness services) apply a rule based tagging system to connect the sensor data with appropriate human activities. A unified labeling method provides a flexible way to handle sensor data with different labeling system.
HDLS
CORTX DOF
Cortx_dof is a DoF integration into Seagate's CORTX server in Python. For managing data, it uses boto3. Cortx_dof helps to integrate PyTorch, TensorFlow and Elasticsearch into CORTX at the same time with the functionality of DoF. The main functions of cortx_dof are: uploading with S3, searching with Elasticsearch, downloading with S3.
Cortx_dicom is a dicom integration into Seagate's CORTX server in Python. Handling dicom files is based on pydicom. It has 5 major functionalities: collecting metadata (cortx_dicom and pydicom), managing data (S3), searching (Elasticsearch), warning about potential legal and data risks (cortx_dicom), removing protected health information from dicom (pydicom). We already integrated 4840 dicom tags into our system as the base of the labeling method. For the legal parts, we implemented the warning features for 256 potentially risky tags.
CORTX Dicom
InfinityBatch
Infinitybatch is an open source solution for PyTorch that helps deep learning developers to train with bigger batch size than it could be loaded into GPU RAM through a normal PyTorch train loop. The core concept of the idea comes from the fact that GPU time is expensive and the usage of own GPU cluster or a cloud based GPU service has to be optimized to be cost efficient. Furthermore, developers and researchers regularly have limited access to GPU. However, CPU based training mostly allows higher batches than a normal GPU could provide, it is much slower. Infinitybatch helps to use GPU during training with bigger batch size thanks to the special unloading and uploading process that manages the GPU RAM to avoid memory overrun.
Synthetic NF1 is a generative method to solve MRI data shortage in the field of neurofibromatosis type 1. Training AI models to identify a rare disease is really hard due to the lack of training data. Creating synthetic MRI images solves this problem. The first phase of the project is creating networks to make realistic MRI images. The second phase of the project is fully controlling the image processing by selecting the size, location and type of tumors. The third phase of the project is creating realistic high resolution synthetic MRI images. This project is in the first phase.
Synthetic NF1
Autonomous Platform
The goal of this project is to create an autonomous rubber track platform that navigates based on visual and sensor data. The ultimate goal is navigating the platform based on visual data only. The platform drives itself on any terrain. A platform like this can be used in agriculture, or as a military device or on the surface of a new planet.
Software to measure the footprints of deep learning models at training, testing and evaluating to reduce energy consumption and carbon footprints. This helps to fight against global warming and wasting resources. It is useful on Earth but it can help to plan long-range space travels as well.
GreenOps
logos
Logos is a real-time cooperative logging and log management system with flexible access control, custom filtering and high security to provide better contribution, safer communication between the mission participants and data flow between humans and autonomous sensors or unmanned vehicles. It can be used as a yet-another logging platform or it helps data transmission between different parties like companies, agencies or fighting corps on the field. Logos uses communication standards that ensure high portability between operating systems and programming languages.
This project consists of two different parts: recognizing human faces and identifying them. Both can work alone or in a chain. Identification can be a total ID recognition or just a semi-identification where each face gives a random number. This project can be used as a security system to let or deny access in protected areas or a system to track movement of people in protected areas. Since the nature of this project relates to legal risks and issues, this project is not open to the public.
FRPI