Back to blog
ai-first /software-house /kubernetes /rust

VertexHub: AI‑first software house with 19 products

VertexHub 13 de julho de 2001 5 min

VertexHub was born as a Brazilian software house that puts artificial intelligence at the heart of every line of code. Since 2025, we operate an ecosystem of 19 digital products – some already in production, others in development or lab – and we offer the Build Together program to build tailor‑made solutions using the same stack, standards and culture that support our own services.

Who we are

  • AI‑first and agentic‑first: AI is not an optional feature we add at the end of a project. From the first commit, our systems are designed to consume language models, vectors and autonomous agents as structural components.
  • Hybrid business model: we maintain 19 proprietary products while developing custom software for other companies.
  • Geographic base: we operate from São Paulo and serve clients that need compliance with local regulations (LGPD, ICP‑Brasil, MP 2.200‑2/2001, etc.).

Architecture and stack

Our technology stack reflects the need for performance, scalability and security:

Layer Technology Reason for choice
Backend Rust (Axum, Actix) • Go (Echo, chi, pgx) Low‑latency languages with high memory control, facilitating integration with AI models and critical workloads.
Frontend SvelteKit 5 / Svelte 5 • TypeScript Compiles to lightweight bundles, ideal for responsive applications that consume real‑time AI APIs.
Database PostgreSQL 17 Advanced support for vector types, extensions and robust transaction control.
Orchestration Kubernetes (k3s) + Kata Containers Micro‑VM isolation for workloads that require secure execution of external models.
Network Envoy Gateway • Cilium CNI Layer 7 load balancing and identity‑based network policies, essential for agent communication.
AI/ML Qdrant (vector DB) • Ollama, Claude, Groq • MCP (Model Context Protocol) Embedding storage, local or API inference, and a standard protocol for context exchange between agents.

These components are reused across all products and Build Together projects, ensuring quality consistency and reducing delivery time.

Production products

Vertex ChatSense

  • What it delivers: an omnichannel customer service platform that aggregates WhatsApp, Instagram, Messenger, Telegram, TikTok and email into a single inbox.
  • Integrated AI: language agents with augmented document retrieval (RAG) that answer queries, route tickets and execute automations.
  • Stack: Go 1.26 (Echo v4 + pgx), SvelteKit 2, PostgreSQL 17.
  • URL: https://chatsense.app

Atendis

  • Audience: self‑employed professionals and small businesses in the beauty, aesthetics and health sectors.
  • Key features: public online scheduling, service‑based margin engine, WhatsApp reminders and conversational AI for appointment booking.
  • Stack: Go, SvelteKit, PostgreSQL.
  • URL: https://atendis.app

Praxia

  • Focus: telemedicine for doctors, psychologists and nutritionists.
  • Compliance: encrypted electronic records (LGPD), digital prescriptions with ICP‑Brasil and recurring billing integrated with Stripe/Asaas + NFS‑e.
  • Technology: WebRTC with real‑time signaling, Go backend, SvelteKit frontend.
  • URL: https://praxiaapp.com.br

Products in development

KandleSmart

  • Goal: provide real‑time financial intelligence for B3 traders.
  • Differentiators: institutional real‑time feed, AI pattern detection and algorithmic signals, professional terminal with Mini Dollar and Mini Index data.
  • Stack: Go, SvelteKit, PostgreSQL, WAConnect (internal Rust library).
  • URL: https://kandlesmart.com.br

EduTechMaster

  • Proposal: white‑label educational platform with AI that guarantees source citation, automatic intervention for dropout risk and legal compliance (CDC, Decree 7.962, LGPD).
  • Advanced features: forensic watermark on HLS videos, AI tutor that responds only with previously cited content.
  • Stack: Go, Rust, Svelte 5, PostgreSQL, NATS.
  • URL: https://vertexhub.ai/projetos

SignID Brasil

  • Service: digital signature with legal validity, aligned with MP 2.200‑2/2001 and Law 14.063/2020.
  • Strengths: auditable signing flow, integration with management systems and support for ICP‑Brasil digital certificates.

AI‑first and agentic‑first approach

  1. AI planning from the start: when defining the architecture, we already choose language models, embeddings and RAG strategies that meet latency and accuracy requirements.
  2. Agents as components: each micro‑service can expose an “agent endpoint” that accepts structured prompts and returns responses or actions, enabling autonomous collaboration between system parts.
  3. MCP – Model Context Protocol: we standardize context exchange (prompt, history, embeddings) between services, avoiding “prompt leakage” and facilitating model reuse.
  4. AI observability: inference logs, token‑cost metrics and drift monitoring are collected alongside other performance indicators, allowing continuous adjustments without downtime.

Build Together – custom‑demand projects

The Build Together program follows three well‑defined stages:

  1. Alignment – we understand the client’s real problem, map business requirements and identify automation opportunities with AI.
  2. Architecture – we define the solution, technology stack, scope and delivery milestones, using the same standards as our proprietary products.
  3. Co‑construction – incremental deliveries with code review, staging environments and automated CI/CD pipelines. The team that builds the project is the same team that maintains our production products, guaranteeing knowledge transfer and quality.

We do not work with fixed price tables; investment is adjusted to the scope defined in the architecture phase. This flexibility allows us to serve everything from startups needing an MVP to large corporations seeking migration to a native AI stack.

Conclusion

VertexHub demonstrates in practice that building digital products with integrated AI is not an isolated experiment but a sustainable development model. Our products – ChatSense, Atendis, Praxia – are already in production, while KandleSmart, EduTechMaster and SignID Brasil are moving toward market, all following the same discipline of architecture, observability and security.

If your company needs a solution that combines the robustness of a modern stack with cognitive agents from the first commit, the Build Together program is ready to turn the idea into delivered code. Learn more, submit your challenge and start co‑building with us at /projetos.

All articles