Wand.ai is an AI-powered company developing solutions for business users, data scientists, and analysts that help to solve business tasks and get more of the possessed data. Their main product is a self-service AI platform that requires no coding knowledge to work with. The product is more than interesting, especially in developing the custom AI solution based on each client’s case of data sources attached and business goals selected.
When Wand.ai came to us, the product idea wasn’t even in the air. We got a request for full-scale consulting for a future project yet to be developed. After a small talk, we discovered the project's conditions and “landscape,” like what it has to do and what loads will be handled.
Still, we understand that the research phase will be more than massive because the client even asked to investigate technologies and tools for future project implementation. So, what input we had before the launch of the consulting service:
- The implemented project has to be multi-tenant and 100% underlayer platform-agnostic.
- The main task it solves is custom ETL processes that can simultaneously handle the load from multiple clients with no limitations to data volumes. Another nuance we had to consider is that the data will come in different formats and should be prepared for AI processing and MLOps processes on production.
- The solution has to be secure, scalable, and fast. The first-class tool, and no less than that.
That was enough for us to start. With a team of Principal Architects, we started providing our consulting services.
The process of consulting delivery
The whole process of any consulting delivery consists of the following steps:
- Definitions of requirements
- Deep research and modeling
— Collecting data and processing it according to the goals of the research
— Selection of the indicators and metrics of the infrastructure that demonstrate the state of the project model and its functionality features
— Testing of the model with the load tests. We tested certain elements of architecture and conducted benchmark tests to compare and choose the most suitable instruments.
- Demo presentations of intermediate results, model review.
If everything is approved, let’s dive deeper!
- Architecture consulting part: Building up the vision of the project’s architecture
— Knowledge database created based on conducted tests
— Guide base for developers regarding each aspect of the project
— Roadmap + development plan
- Numerous demo presentations with the clients’ team, standup meetings, and Q&A sessions.
Our team worked in the closest connection with the internal team so that they could ask us about any details regarding the project anytime.
Consulting services that we completed for Wand.ai
While following the service delivery process, we started by creating the helicopter view of all requirements we have to align with and building up an action plan. Like archeologists, we won’t find the treasures we’re looking for without a direction and definition of the working fields. So we started with clarification of our goals and estimations with a client.
The main milestones we’ve been through in the framework of consulting were the following:
- Business and technical idea validation
We check if it’s generally possible to create such a solution using ready-made components, as we require fast delivery. We also hand-pick the technologies, tools, and solutions that suit the product and will be used for development. While doing this, we don’t forget about matching the business requirements and resource limitations.
- Creation of the overall product vision
We form and describe the product from A to Z based on the input demands. The result is the architecture with described layers and blocks and an explanation of their connections.
- The core functionality analysis and mapping
At this stage, we atomize and connect the bits of functionality, ensuring nothing is left behind. We check that our offered solution is efficient, has a minimum of code, and is still editable, controllable, and easy-scalable. Thus, we build an architecture for the future solution, choosing the best options based on 100% of the client’s requirements, capacities, and available resources.
- The development plan for the project
We prepare a whole list of components with direct explanations of what tools/technologies/frameworks to use here or there and a roadmap of tasks for developers. With this guidance alongside other documentation we develop, the process of creation of the product will take significantly less time than without prepared architecture.
The time it took to provide the whole range of our consulting services was less than 2 months, with continuous delivery of each part of our work. This is an investment in the further smooth development of the project with an approved idea and architecture that 100% fits the need. With all these preparation, the project has all chances to develop correctly, be competitive in the market, and show the best possible results for the end user.
The value generated by Dysnix Architecture consulting services
After we presented the architecture schema and many other docs and handled all the events connected to the presentation of the results, it was our turn to make conclusions:
- We clarify the business idea and make it defined. Now the client knows for sure what kind of product they will develop.
- The architecture we’ve developed fits all the current and expected needs of the client. The project gets the direct way of realization and technical basis for it.
Our solution is an efficient way to merge any type of data with a minimal amount of code, as we created a seamless pipeline workflow that meets the high-security demands. The architecture we modeled is stable and high-available. The included continuous deployment feature guarantees zero downtime, even during huge core updates.
Also, the developed solution supports ML models' training and serving without much human intervention. We planned how to launch the full-scaled MLOps in production using the hand-picked technologies for this.
Another thing that should be mentioned is that the whole solution is automatically scaled. So the load balancing functions are also included.
- Each project detail is described in the actionable documentation that will be used in the development process. Each tool has a description and guide for application at every stage of development.
- Our model and testing prevent possible problems with the project that might be caused by low-quality design. We included our expertise in building and developing such projects in the planning stage, so many pitfalls, unnecessary expenses, and problems are avoided before the appearance.
So that’s how with the help of Dysnix consulting, Wand.ai started to develop a competitive product with confidence that no startup can have.