Consulting and Support for 3D Configurator Development
We support in-house development teams or individual developers in bringing their 3D and CPQ configurators to production. We offer a modular, scalable path tailored to your needs, starting with an audit and initial analysis of the codebase and continuing with architectural consulting sessions, development support and code review.
We start from different points of departure (React projects, prototypes built with AI on Lovable, Replit or with Claude Code and Codex) to reach the same result: a secure, operational configurator, integrable with your management software and CRM and useful for your work in sectors such as manufacturing, furniture, nautical, medical devices and much more.
Building in-house with AI
Today's coding agents are extremely powerful. It is therefore natural for many companies to experiment with them, and to do so in a sector structurally underserved by complete SaaS solutions, such as 3D configurators and CPQ quoting tools.
Whether it's a person without a specific technical background or the IT team that usually manages other software, the start is always a "honeymoon phase". Coding agents work very well starting from scratch, and the code always compiles.
After a while, though, two problems come up that can certainly be solved by a person, even a non-technical one with the will to learn and the time available, but that benefit greatly from a conscious approach and a certain amount of experience in the field:
- Codebase size: past a certain threshold (estimated a few months ago at around 10,000 lines of code, today I think it's higher, though I plan to do some research on this) even coding agents struggle more, and every new feature drags along more weight
- From generic to specific: the start was necessarily fairly "standard", while the result you want to achieve is extremely specific. The move from generic software that renders 3D to your company's configurator, tailored to your products, your customers and your sales team, is far more demanding than the move from nothing to a demo
On top of this there is always the nagging issue of security: the level of attention cybersecurity requires today has increased significantly. The very same coding agents that can help you and us develop can also be used by malicious actors to build exploits in fields where, years ago, it was not reasonable for them to operate: manufacturing companies are certainly one of these.
Professional, specialized support for in-house development can be the ideal solution to keep full control over the software, save significantly and, at the same time, benefit from well-designed, scalable, maintainable and secure software. Moreover, the consulting investment pays for itself quickly through the hours saved solving problems that are standard in this vertical.
How development support solves the problems of AI-built software
The two problems you run into after the initial honeymoon phase share the same root cause: AI models are trained on many millions of lines of code across thousands of different uses, and their objective in each iteration is to build what you ask for and make it compile. The bigger picture (the context) is collateral to this objective: managing it correctly is exactly what current research is about, and exactly what every professional software engineer wrestles with every day.
The problem is that you don't need generic software, but extremely customized software, for which AI has few examples of comparable optimizations. Because of this, it very often tends to "over-engineer" and pick solutions that work for the feature you're asking for right now, but that may create problems down the line. This creates both the wall you hit at 10,000 or 20,000 lines, and the difficulty of reaching behavior that is truly perfect for your product and your company.
At the same time, this is kept under control with a clean, precise and as simple as possible code architecture. This architecture is achievable with discipline and by taking the time needed to design component architecture, the contracts between user choices, the bill of materials and 3D rendering (where present). It requires constantly revisiting past choices and removing over-engineering, especially in the fundamental choices around data structure and responsibility boundaries between components and sections of code. And a certain amount of experience on similar projects helps with this.
The path
It all starts with a free introductory call to get to know each other and dig into your project's goals.
Following the introductory call, we can provide a feasibility estimate and duration for the support path.
After that, our work is structured in two phases:
Audit and analysis of the codebase's initial state
We access the codebase (preferably via a GitHub repository) and analyze the state of the project, with particular attention to:
- Architecture and maintainability: code organization, duplication, dependencies, ability to evolve without regressions
- Data schema and sources of truth: catalog management, external data sourcing, any inconsistencies between states, whether or not state needs to be centralized, correct separation between sources of truth for the different data the software needs to work
- Security: authentication, data access policies, key and secret management, API exposure
- Performance: analysis of the current code structure against performance requirements (especially for 3D)
- Pre-feasibility of further developments: identifying any mismatches between the current state of the data schema and codebase and the future features to be implemented
The result of this phase is a concise, point-by-point report, accompanied by an explanatory call.
Consulting and development support sessions
Indicative duration of 2 hours, with specific, agreed-upon goals, such as:
- high-level strategic and architectural decisions, such as drafting a development plan with clear milestones, refining the general scope, caching strategies for sourcing data via API from your management software or CRM, etc
- producing executable high-level spec files (directly or in steps) with coding agents (such as, for example, the architecture of the scene's components and the split of responsibilities between components, or the contract between the state machine, the scene and the bill of materials)
- producing the specifications to hand off to whoever does the modeling, or to use directly for 3D modeling or optimization
- final or intermediate code review
Every session includes a review of the changes made since the previous session.
Sectors
We work with configurators and quoting tools for manufacturing, furniture and kitchens, windows and technical closures, nautical, medical devices and industrial machinery. The projects we've built give a concrete sense of the complexity handled: catalogs with thousands of codes, compatibility rules, bills of materials, ERP and CRM integrations.
Frequently asked questions
Can a configurator built with AI go into production?
Yes, as long as it passes the checks a demo doesn't face: secure data access, a data schema that can hold up under the real catalog, isolated and verifiable pricing and bill-of-materials logic, and code that stays maintainable over time. The audit exists precisely to establish the distance from this standard and the path to close it.
What technologies do you work with?
Modern JavaScript stack: React and Next.js, Three.js and React Three Fiber for 3D, Supabase and similar cloud services for data and authentication. We also support prototypes born on Lovable and Replit, or generated with Claude Code, Codex and Cursor, which typically rely on these same technologies.
How is code confidentiality handled?
Read-only access to the repository, by invitation and revocable at any time; we're available to sign an NDA before access. The code we review is never reused.