Start-ups need to test hypotheses quickly. However, to launch an MVP, infrastructure, and development tools are required.
It is possible to run on your infrastructure, but often, teams do not have enough computing power, and they choose the Cloud. Cloud tools for development and working with data allow you to quickly design a solution architecture and deploy it on the Cloud, and as the project grows, add computing power without purchasing equipment.
In this article, we tell you how the Cloud helps solve 3 main problems that start-ups face when creating and scaling a product.
Task 1. Quickly Create And Test An MVP
Problem. To create a successful project, the team tests hypotheses, creates prototypes of solutions, tests them, and selects the best option. Speed is essential to a start-up: a quick launch allows you to assess the potential of an idea and adjust plans.
Solution. The Cloud has a set of development tools that can be used to design and test a prototype quickly. If the MVP version is successful, then a full-fledged solution can be deployed in the Cloud. The idea didn’t work – you can rebuild the prototype or suspend the project by turning off the power.
Real case. Using the Cloud platform, scientists have created a neural network that helps monitor colonies of Red Book pelicans. Counting birds manually is a labour-intensive process, so it was automated using a computer vision-based model. The neural network analyses images from a quadcopter and determines the number of birds in the population in 30 milliseconds; this took 7 days manually.
To quickly and cheaply test the idea and develop an MVP, we used the Cloud ML Platform service. The team has now completed testing the neural network and plans to move on to developing a full-fledged solution.
Task 2: Quickly Gain Access To A Set Of Tools For Creating A Product
Problem. When a team works on its IT infrastructure, it has to buy software or use Open-Source solutions. This requires additional costs for software, its deployment, and administration.
Solution. On the cloud platform, the tools are already configured and integrated. For example, Cloud offers more than 30 services for developing and working with data, which can be deployed in a couple of minutes.
There are databases in the form of a ready-made service and managed Kubernetes with an autoscaling function – as the load grows, it automatically adds new nodes. Cloud ML Platform is also available – a pre-configured solution for working with ML models (the service is built on the basis of JupyterHub and MLflow). The team does not have to wait: select the tools and the required configuration, and you can immediately begin development.
Real case. The lead generation agency Likwid has created a service to automate the search for clients for B2B companies. The product uses NLP to analyze the semantic core and, using algorithms, searches for websites of companies that match the portrait of the customer’s target audience.
The service is based on learning neural networks, so a lot of resources are spent on support and development. For fast data processing on the Cloud platform, Kubernetes and Cloud DBMSs are used: PostgreSQL and MongoDB. The system was deployed in the Cloud in just a week and a half.
Objective 3. Maintain Growth And Scale A Successful Project
Problem. When a project develops rapidly, computing resources begin to run out. If, at the same time, he works on his IT infrastructure, then he has to buy additional equipment, and if there is a lack of money, he artificially slows down the growth of the project; that is, he loses revenue.
Solution. In the Cloud, you can scale your computing power in a couple of clicks.
In this case, you pay according to the Pay-as-you-go model – only for the resources actually used. For example, you are conducting a marketing campaign, and the site has more visitors. To prevent it from falling under load, you can temporarily connect additional computing power. After the promotion ends, you turn off unnecessary resources, and the fee is reduced.
Real case. The startup MedTech AI is working on projects related to the processing of medical images and documents. The company uses language models of the GPT family (1–3), which require enormous computing power to operate. Therefore, we were looking for a solution that could cope with high loads, provide stable performance, and the ability to scale in case of launching new projects.
On the Cloud platform, MedTech AI uses graphics accelerators (GPUs) for image recognition. Thanks to this, it was possible to increase the speed of training ML models – from two and a half weeks to several days.