The MVP Architecture
As a Gen AI Apps design and development provider, Arrays Lab creates the architecture blueprints per a use-case. You can start with a blueprint for any of our pre-built use-cases or we’ll design your unique custom app. You can book a 2-hour design consultation and we will design it for you during the session. We will work together with you to define your project scope, feasibility, timelines, and resource requirements.
The architecture establishes a base to start coding, get the app up & running very quickly and for further scaling. It provides you with a roadmap for your app development and includes the following parts:
Output Expectations
Formulating the specific outcomes and deliverables you can expect from the app: this includes the type of results, format of outputs, and how they align with your use-case (business) goals.
the App Modeling: its high-level design
Modeling the core components and their relations within the gen AI (software) app: we’ll create the Entity-Relationship (ER) Diagram that visually represents all key elements such as users and objects. This diagram will serve as a foundation for understanding the app’s structure.
The Usage Volume Expectations and Costs
An estimate of the expected usage volumes and associated costs: this can include projections for API calls volume, data storage requirements, and processing needs, helping you plan for resource allocation and budgeting.
The Models’ Evaluation
We’ll assess (preliminary) various AI models suitable for your use case, comparing their performance, accuracy, and efficiency. This evaluation will help project the final selection of the most appropriate model(s) for your specific needs and define whether it’s required to further fine-tune a model.
The Workflow Design, Tools and Third Parties
The workflow: a user flow and stages, outlining and detailing how the requests and data move through the system. We’ll specify the tools and technologies to be used at each stage, as well as the third-party services or APIs that will be integrated.
The Processing Environment Setup
The requirements for the processing: a choice of a future (cloud) provider may depend on specific AI services offered, and pricing. The processing environment will be set up to handle the computational demands of AI model inference, with GPU instances available for tasks like image generation or complex NL processing.
Scaling Strategy
We’ll provide a preliminary plan for how the app can work as an MVP, and how it can graduate to the production environment, grow and handle increased demand. This includes strategies for rapid development and ensuring that your app can maintain performance as user numbers and data volumes increase.
Bonus 1: 2 tech Stacks
Usually we do 2 versions of the architecture with a tech stack to apply for each. We start with an MVP blueprint and the best tech stack to make it happen really fast. We add some projections for a production-grade version of the app with a tech stack to use.
Bonus 2: Execution Planning
Timeline, team and tools, including a timeline for MVP, rest run, and scaling.
Explore Our Architecture Solutions
Chose the most suitable options.
A Design Consultation & a Workflow
$400
A (Pre-built App) “as-is” Installation
$4,000
An Installation with Fine-Tuning
$6,000
Custom Solutions: Per Month
$3,000
A Data Suite: training synthetic data
$24,000
Example Downloads
Get in touch to download examples to make informed decisions.
The MVP Architecture Example
Get in touch to download an (“proof-of-concept”) MVP app design blueprint example.
The Model Evaluation Example
:a live example of a model evaluation and (process) how we determine if it should be fine-tuned.
The Model Fine-Tuning Example
You can chose example for an image model fine-tuning, transformers, or embeddings (RAG).