Does it interest you if the medicines you take are real, authentic products or fake, forged ones? It sounds like a radical question, perhaps, but forgeries are a huge problem worldwide and need effective solutions as soon as possible.
Here at Spectral Engines, we’ve developed a material sensing solution that will save both time and investment at the setup and operation phases of building, for example, an application for detecting forgeries. These savings are one of a kind in the world of spectral sensing: they could be several tens of thousands of euros per year. But saving money is not the only benefit of our solution: your end-product will also be implemented easily and securely.
The solution comprises a unique material sensing device based on scientific proven near-infrared spectroscopy, and the data processing is done using the latest tools in the field of machine learning. As a result, this solution opens new business models in applications where highly portable, cost-efficient, and fast material scanning is needed.
As operational costs are charged monthly for each device, business models where you pay as you go are also possible. We’re proud to say that we’ve managed to minimize the costs by moving most of the intelligence to a cloud platform. Amazon AWS cloud infrastructure is trusted by many blue-chip companies, and also by us. The cloud is based on the micro-service architecture with Dockers to make easy scaling of computing resources, load-balancing, and high-availability possible. The primary building blocks of our material sensing solution are mass-producible sensing hardware, cloud infrastructure for the fleet management, and application-specific machine learning models for high-speed data processing.
The development process
The implementation of our customer’s use-case starts with a measurement campaign where presentative data on real substances get gathered. This will form a basis for the machine learning model creation to get a well-performing application in the end. It’s crucial to have a grip on both the chemistry of the application and the optical performance of the sensing hardware. This will enable us to create accurate models that can be deployed to the whole fleet of sensors. We use machine learning models instead of traditional chemometrics to get control of three sources of uncertainties. Firstly, disturbances caused by a user or an operator who is doing the measurements in the field. Second, sensor to sensor variation that is normal for cost-efficient sensors and third, sample to sample variation, because there is always material inhomogeneity and process variation. The machine learning models are then deployed to the cloud platform and tada! It’s then usable right away.
Version control and release testing are, of course parts of the process, as they ensure high availability for all users. All of this makes it possible to deploy a single model for the whole sensor fleet, which is an essential feature when enabling low installation and operation costs.
The centrally managed cloud enables data analytics, trend monitoring, preventive maintenance, and improved customer support. REST API connections make the integration to the customer’s own IT system possible. These together enable history data gathering to customer’s dashboard or quality monitoring systems. Our distributed microservices are monitored using the Elastic, which provides real-time information on the load and computing time of our cloud system. The communication between these services are encrypted with TLS, and an additional layer of security is made possible by using the Cloudflare service. The data is transported securely, and the system is harnessed against DDoS attacks.
The benefits of our solution
The material sensing solution by Spectral Engines opens radically new business models with different kinds of pricing models, starting from one-time payment to a monthly subscription and even leasing. In comparison, business hardware sales and service agreements are traditionally pretty typical. A fascinating feature is that the new applications and updates to the existing applications can be deployed fast to the whole sensor fleet so that the end-user does not see any service interruptions. This enables upselling and cross-selling.
Also, scalability makes starting with a small fleet, and then scaling up to tens of thousands of sensors happens possible without any effort. The NIRONE Scanner is the basis of all cloud products at Spectral Engines. We utilize it to realize two of our own products: TactiScan for drug screening and PurplPRO for cannabis potency. Applications based on the NIRONE Scanner can be maintained by Spectral Engines or a distributor, depending on the business model. The solution is very flexible for different kinds of applications and business needs.
In a nutshell, the NIRONE Scanner is a novel and secure cloud-based material sensing solution for distributed applications. This opens new business possibilities - and shortens market entry time. Sounds good, right?
By Uula Kantojärvi, CTO, Spectral Engines
