KASCO TECHNOLOGY

Kasco R&D is a company specializing in the development of machine learning solutions, we provide full stack solutions that encompass algorithms, software and IoT devices. 

Headquartered in Campinas, São Paulo, with more than 17 employees, Kasco maintains a highly qualified research and development team and strategic partnerships for the delivery of innovative products. Kasco’s diversified customer base includes energy concessionaires, multinational consumer goods, security companies, agro-business multinationals, and research and development promotion institutions. Currently, she is a pioneer in the development of artificial intelligence solutions for different industries. Among the main solutions developed, we can mention: The MVI (Multi Vision Inspection), referring to the use of Deep-Learning for inspection of electric power lines using land vehicles and drones. 

FaceId is another solution that uses the state of the art in Deep-Learnging algorithms for facial detection and recognition, being a complete facial recognition platform, which integrates detection system, integration API’s and management system for the generation of accurate reports. Kasco A.I (KAI) is a proprietary structure that integrates a dataset store (Kasco Image Store), a store of custom Deep-Learning algorithms (Deep+) and Integration API’s. 

KAI provides critical support for the training of algorithms and product improvement processes allowing the company to build a production line of semi-automatic algorithms that constantly improves and becomes more autonomous over time. 

As a differential, Kasco has the ability to develop new neural networks with less human effort and shorter delivery time, constantly seeking the best results. Kasco also has KVI (Kasco Visual Inspection) as a solution for quality inspection of products in industrial production lines. 

Using the Deep-Learning frameworks themselves for the inspection of all products, it can identify flaws in packaging, wrappers, boxes, lids, labels in ultra fast lines, analyzing flaws in different product views and different SKU’s.

Team, Technologies and Differentials

Kasco R&D has developed over the years a high performance team with the ability to operate in different areas of knowledge. This team consists of a combination of researchers with more than 15 years of experience in the development of projects and products, with newly graduated developers and interns, as well as executives and consultants, thus combining different ideas and visions throughout the development of the product. 

Among the technologies and knowledge, Kasco has specialized in integrating a diversity of proprietary technologies, frameworks and platforms open to the evolution of its products, always customizing and developing new modules to meet customer demands and better use of hardware resources. Regarding the frameworks for artificial intelligence applications, we can mention:

Artificial Intelligence

CUDA, Deep Learning SDK, Jetpack SDK, MXNet, Pytorch, TensorFlow, TensorRT, TRTIS, TLT.

GPUS NVIDIA:GeForce, Tesla T4,NVIDIA V100, NVIDIA DRIVE, Jetson TX2, XAVIER AGX.

Algorithms: TTFNET, Yolo,  Alexnet, MaskRCNN, CascadeRCNN,

Software Projects

Kasco has its own Framework to accelerate deliveries of its applications. The Genus 2.0 framework aims to streamline the development of web systems, including both front-end and back-end and consists of a set of code, classes, templates, and ready-made functions that run generic roles, allowing developers to focus on solving the customer problem itself and to reuse interesting modules between projects, without having to rewrite already tested and functional code snippets. 

Genus in particular already has a skeleton and an organization of files that makes the task of adding new pages and features simpler and in its purest version already has, among others, the following components: 

  • Login screen Permissions system for each user and access groups (defining the pages that each can see /edit)
  • User administration functions (registration, editing and deletion) Basic integration with database (add and edit users)
  • Screens with access logs, security and errors for the administrator
  • Skeleton of pages, with sidebar containing menu and information on the current page
  • Responsive design, adapting to different screen sizes, including mobile 
  • Configuration file to add new routes and pages easily

Node.js and Express.js

The Genus 2.0 backend was developed using Node.js, a JavaScript interpreter that runs on the server. It is lightweight, supported by several companies and uses a package manager called NPM (node package manager) that manages both application development and publishing dependencies, containing a large number of modules available for use. 

Another advantage of using Node.js on the server is that the programming language used in the backend is the same as the front end, so development is simpler, and it is not necessary to empower developers in another language like PHP. In addition to Node.js, Genus 2.0 also uses Express.js, a framework for building applications and APIs. It is considered the default framework for working with Node.js. 

Express is used to define the routes that will be consumed and develop middleware (functions that are executed sequentially on the request passed by the client) thus speeding up the development of server code.

MongoDB and Mongoose

 

MongoDB is a non-relational database (NoSQL) document-oriented. It stores documents in BSON format, similar to JSON (JavaScript Object Notation). It is currently the most widely used non-relational database. Unlike relational databases that store data in tables with defined schemas, MondoDB allows great flexibility by not requiring all new entries to have exactly the same format. 

In addition, disk, MongoDB has the possibility of horizontal scaling, in which it is possible to expand the database by adding a new database, while for relational databases would require a hardware improvement (vertical scalability). Together with MongoDB Genus 2.0 uses mongoose. 

Mongoose is a library for Node.js that allows you to model the data and map it from the database structure to JavaScript objects so that they can be used by the application being developed.

Hardware Projects

Kasco R&D has in its history the design and development of some Hardware equipment, and can highlight: PCie Octotuner ISDB-T Board, DUAL TUNER ISDB-T Board, a miniPC UHD-Player with GPUs for A.I., physical structure with cameras for thermal and optical inspection, among others. 

The projects are always developed considering the smallest dimension, the required amount of layers, the cost, the quality of components and welds, the thermal efficiency, durability, the lowest energy consumption and usability. 

The hardware development steps involve the 3D design of the boards, layout design, component list, gerber files, prototype production, local testing, scale production, and factory product testing prior to shipment to the end customer.

Development of Drivers, API and SDKs

For the control and operation of the developed hardware, Kasco develops its own drivers and SDK (Software Development Kit) to ensure the compatibility of its equipment on various operating systems such as different versions of Linux, Windows, Chrome, among others.

Programming Languages, Protocols and Webservices

Seeking the maximum performance of its systems and applications, Kasco constantly seeks to identify the most appropriate development languages for the development and future support of its products, among which we can mention: CUDA, Python, C #, C ++, Java, Javascript, Php , Typescript, Ruby, Swift, Objective C, among others. 

Regarding the choice of protocols, for each application the appropriate protocol is sought, seeking to reduce response times and latency in data communication. Among the various protocols analyzed are: 

UDP, HTTP, SMTP. CORBA, XML-RPC,

WEBSERVICES: SOAP, WSDL, UDDI 

MARKING LANGUAGES: XML, JSON