Our Platforms use only open source technology stack

We use top tools to facilitate efficient development activities.

• Jetbrain Team City for CI/CD
• VSCode and Jetbrain IDE for software development work
• Slack for task management and communications
• Github and Git GUI Client for code management
• Matlab Toolbox

We work on highly scalable machine learning algorithms capable of dealing with big data using the AutoML frameworks. Our portfolio consists of in house AutoML library based on Python/TensorFlow and also commercial tools like Azure, Google AutoML. Our support tools includes a good library of data cleaning, data integration, data transformation, data reduction, and data discretization and quantization functions before the data are fed to Machine Learning models. We use commonly available AutoML tools to automate End to End Machine Learning workflow that includes process of Data Gathering, Data Preparation, Model Identification and Building, Model Validation, Orchestrate flow of data to multiple ML models during training and testing and finally Model Deployment.

Metadata is used to describe a scenario. It acts as the glue between layer 1 and the subsequent layer.

1. Enables a common schema for communication across layers
2. The platform provides a library of metadata templates that can be used as-is for various use cases across functional domains.
3. Metadata captures event attributes including event type, timing, location and source of event generation.
4. The Metadata forms the basis for Visualization Analytics in subsequent layers.

Reasonably the most powerful applications of machine deep learning in computer vision technology is object detection. We use Inference layers using pre-trained heterogeneous concurrent deep neural networks that are retrained for our use cases. Our products use various networks including image classification, such as ResNet50 and Mobilenet, computer vision use cases of object detection/classification and segmentation.

Platform built on Multi Layer Architecture with each layer decoupled and integrated using Kafka Message Broker.

1. Layer 1 contains heterogeneous concurrent deep neural network with capability to convert Metadata to message instances and dynamic stream management.
2. Layer 2 contains large scale Kafka broker services for Metadata processing streamed from Layer 1.
3. Layer 3 contains Analytic Services including unstructured data store and Visualization.

Plugin frameworks based on C++ and Python allows us to

• Add new models to derive additional functionalities without impacting existing flows.
• Integrate the outcome of the models to other systems like CRM, Analytics store etc.

Our dashboard, visualizations, and predictive applications uses a multitude of technologies which are simple and functional. using JSON technologies like Vega, AntV L7 for advanced Visualization functions.