Build Together: how VertexHub develops custom software

Developing custom software can be as complex as the idea that motivated it. At VertexHub, we turn this complexity into a structured flow, maintaining the same discipline we use in our 19 proprietary products. The Build Together program was designed for clients and the engineering team to work side by side, from problem understanding to delivery of tested, production‑ready code.

What is Build Together#

Build Together is VertexHub's official arm for custom projects. It follows a clear three‑stage process, without fixed price tables: everything is defined based on the client's real needs and the architecture that best serves them. The same team that delivers Vertex ChatSense, Atendis, Praxia and other products applies here the same stack (Rust, Go, Svelte, PostgreSQL, Kubernetes) and the same quality, security and observability standards.

Stage 1 – Alignment#

The starting point is to understand the client's challenge, not just the list of desired features. In this phase, the product and engineering teams:

1. Maps the problem – detailed conversations with stakeholders to identify critical flows, regulatory constraints (e.g., LGPD or specific health standards) and business goals.
2. Defines success metrics – which results will indicate that the solution is delivering value (response time, conversion rate, churn reduction, etc.).
3. Explores technical feasibility – we preliminarily check whether the required integrations (external APIs, digital signature services, etc.) are compatible with our stack.

The goal is to finish this stage with a requirements document that reflects the real problem and not just a list of “features”.

Stage 2 – Architecture#

With the scope well defined, we move on to designing the solution. This phase includes:

• Architecture design – selection of microservices, use of Kata containers for isolation, definition of entry points (API Gateway with Envoy) and persistence (PostgreSQL 17 or Qdrant when vector search is needed).
• AI planning – how VertexHub's AI‑first posture is incorporated into the project from the first commit. For example, if the product needs a support assistant, we already include an LLM agent with Retrieval‑Augmented Generation (RAG) connected to our context model (MCP).
• Roadmap and milestones – we split development into incremental deliveries, each with a clear scope definition, acceptance criteria and review dates.
• Effort estimation – based on the architecture, we calculate the required effort and present the client with a budget based on the defined milestones, without pre‑defined price tables.

All planning follows the same versioning, code review and CI/CD practices we use in our production products, ensuring consistency and traceability.

Stage 3 – Co‑construction#

The co‑construction phase is where the code actually takes shape. It relies on incremental deliveries, constant reviews and a continuous integration pipeline that includes:

1. Collaborative code review – each pull request is reviewed by at least two engineers, including an AI specialist when model or agent components are present.
2. Staging environments – before releasing to production, the solution runs in an environment that replicates the final infrastructure (k3s, Cilium CNI, Envoy).
3. Automated testing – coverage of unit, integration and, when applicable, performance tests of AI calls.
4. Iterative feedback – the client has access to regular demos, can validate functionalities, adjust flows and suggest improvements before the next cycle.

This incremental approach reduces risk, enables quick fixes and keeps the client always aware of progress. In the end, we deliver a product ready to be operated in production, with documentation, monitoring configured and support for any adjustments.

Why choose VertexHub for your custom project#

• AI‑first from the start – we don’t treat AI as an “add‑on”. If the project requires agents, RAG or document classification, these components are already integrated into the design and code.
• Proven stack – Rust and Go for backend, Svelte for frontend, PostgreSQL 17 for critical data and Kubernetes (k3s) for orchestration. This combination already powers our production products, such as Vertex ChatSense and Praxia.
• Regulatory experience – we have dealt with health requirements (ICP‑Brasil, LGPD) and digital signature (MP 2.200‑2/2001, Lei 14.063/2020) in products like Praxia and SignID Brasil. This background reduces the learning curve in projects that need compliance.
• Quality standards – code reviews, CI/CD, automated tests and monitoring are part of the contract, not optional items.
• Investment transparency – the budget is defined in the architecture phase, based on clear milestones, avoiding surprises along the way.

Next steps#

Building together is not just a slogan; it’s the daily practice of those who put AI, high‑performance engineering and business focus side by side. We are ready to turn your idea into functional code – and, more importantly, into measurable results.