Case Study: Building EmpireVault — From Concept to Scalable SaaS Platform in 90 Days

The Challenge

A B2B software company needed to go from zero to a fully functional, production-ready SaaS platform — fast. The vision was ambitious: a unified customer engagement platform combining CRM, support ticketing, AI-powered live chat, and email campaign management into a single product. The founders had validated the market opportunity but lacked the senior technical leadership to architect, build, and ship a scalable platform within their target window of three months.

The business requirements were demanding. The platform needed to support multi-tenant architecture with per-seat pricing, integrate with major email providers (Google Workspace, Microsoft 365, IMAP), include enterprise-grade security from day one, and be ready for early-access customers by the end of the build cycle. There was no existing codebase — this was a greenfield build.

The Business Case

The customer engagement SaaS market is crowded, but most existing tools force businesses to stitch together separate products for CRM, support tickets, live chat, and email outreach. The founders identified a gap: small and mid-market B2B companies wanted a single platform that could handle the full customer lifecycle — from initial website visit through lead capture, sales pipeline, ongoing support, and re-engagement — without the complexity or cost of enterprise suites like Salesforce or HubSpot.

Speed to market was critical. Early-access signups were already being collected, and a funded competitor was building in the same space. The 90-day timeline wasn’t arbitrary — it was a competitive necessity. Every week of delay risked losing first-mover advantage in a specific vertical.

As the fractional CTO for this engagement, my role was to own the full technical strategy: architecture decisions, technology selection, infrastructure design, development workflow, security posture, and delivery execution.

Technology Choices

Every technology decision was evaluated against three criteria: speed of development, long-term scalability, and operational simplicity. Here’s what we chose and why.

Cloud-Native Kubernetes on Civo

We deployed the entire platform on Civo Kubernetes — a managed K8s service built on K3s that provides fast cluster provisioning (under 90 seconds), predictable pricing, and excellent performance for SaaS workloads. Choosing Civo over AWS EKS or GCP GKE saved significant infrastructure cost during the build phase while providing a clear upgrade path if the platform needed to scale to hypercloud later.

The infrastructure was fully containerized from day one. Every service — the web application, background job workers, AI inference endpoints, email sync workers, and the real-time WebSocket layer for live chat — ran as independent deployments within the cluster. This gave us the ability to scale individual components independently as usage patterns emerged.

Full-Stack Web Application

The application layer was built as a modern full-stack web application with a clear separation between the API backend and the frontend interface. The architecture followed a service-oriented approach where the core platform handled authentication, authorization, and data persistence, while specialized services managed email synchronization, ticket routing, and campaign delivery.

Key architectural decisions included:

• Multi-tenant data isolation — Each customer’s data is logically separated with tenant-scoped queries enforced at the application layer, ensuring no data leakage between accounts
• Event-driven background processing — Email sync, AI scoring, campaign sends, and webhook deliveries all run as asynchronous jobs, keeping the user-facing application responsive
• Real-time WebSocket layer — The Engage live chat widget required sub-second message delivery between website visitors and support agents, handled through a dedicated WebSocket service
• Standards-based authentication — Integration with WorkOS for OAuth 2.0, SSO, SAML, and OIDC — no local password storage, enterprise-ready from launch

AI and Machine Learning Components

AI wasn’t bolted on as an afterthought — it was woven into the core product experience:

• AI-powered ticket triage and routing — Incoming support tickets are automatically classified by category, urgency, and sentiment, then routed to the appropriate team member. This eliminated the manual triage bottleneck that plagues most support workflows
• Smart draft replies — AI-generated response suggestions for support agents, trained on the organization’s knowledge base and previous ticket resolutions. Agents review and send with one click
• Lead scoring — CRM contacts are scored automatically based on engagement signals (email opens, chat interactions, page visits), helping sales teams prioritize outreach
• Conversational AI chat — The Engage widget can operate in AI-assisted mode, handling common questions and qualifying leads before routing to a human agent

All AI inference runs within the Kubernetes cluster as containerized services, keeping latency low and data within the platform’s security boundary.

Enterprise Security from Day One

Having led four consecutive ISO 27001 audits at Honeywell, I brought an enterprise security mindset to the architecture from the start — not as a compliance checkbox, but as a core product differentiator:

• AES-256-GCM encryption at rest using ActiveRecord Encryption for all sensitive fields
• TLS everywhere — all data in transit encrypted, including internal service-to-service communication
• Multi-layer RBAC — role hierarchy, tool-level entitlements, and subscription-based feature gating
• SOC 2 Type II aligned controls — audit logging, access reviews, and incident response procedures designed for eventual certification
• No local passwords — all authentication through WorkOS OAuth/SSO, eliminating an entire class of security vulnerabilities

The 90-Day Timeline

Delivering a production-ready SaaS platform in 90 days required disciplined execution and ruthless prioritization. Here’s how the build unfolded:

Weeks 1–2: Architecture & Foundation

Defined the system architecture, set up the Kubernetes cluster on Civo, established CI/CD pipelines, configured DNS automation and TLS certificate management, and scaffolded the core application with authentication (WorkOS), multi-tenant data models, and the API layer. By the end of week two, we had a deployed skeleton application with working SSO login.

Weeks 3–5: Core Product — CRM & Email

Built the unified inbox with Google Workspace, Microsoft 365, and IMAP sync. Implemented contact and organization management, the sales pipeline with drag-and-drop stages, and the email campaign engine with template management, scheduling, and compliance controls (unsubscribe handling, bounce processing).

Weeks 6–8: Ticketing & Engage Widget

Developed the support ticketing system with AI-powered triage, SLA tracking, escalation rules, and CSAT surveys. Built the Engage embeddable widget — a single-script-tag installation supporting live chat, contact forms, waitlist signups, and newsletter opt-ins with full visitor attribution.

Weeks 9–10: AI Integration & Integrations Platform

Integrated AI components: smart draft replies, lead scoring, ticket classification, and conversational chat. Built the integrations framework supporting OAuth 2.0, SAML, and OIDC connections to third-party services with encrypted credential storage and least-privilege scopes.

Weeks 11–12: Security Hardening, Testing & Launch

Conducted security review and penetration testing. Implemented rate limiting, abuse detection, and monitoring. Built the pricing and subscription management system with Stripe integration. Finalized documentation, onboarding flows, and the early-access program. Deployed to production and opened for early-access customers.

Results

The platform launched on schedule — 90 days from first commit to early-access customers signing in. Key outcomes:

• 5 integrated products shipped in a single platform: CRM, Ticketing, Engage Widget, Email Campaigns, and Integrations
• Enterprise-grade security from day one — no retrofitting needed for B2B sales conversations
• Sub-second live chat delivery through the real-time WebSocket architecture
• AI-powered automation across ticket triage, draft replies, lead scoring, and conversational chat
• Production Kubernetes infrastructure with auto-scaling, health checks, and zero-downtime deployments
• Per-seat pricing starting at $19/seat/month — competitive positioning against fragmented tool stacks costing 3–5x more

Key Takeaways

Building a production SaaS platform in 90 days is achievable, but only with the right combination of experienced technical leadership, disciplined architecture decisions, and ruthless scope management. The critical factors were:

1. Architecture decisions made on day one, not day thirty. Choosing Kubernetes, defining the multi-tenant model, and selecting authentication infrastructure upfront prevented costly rewrites later.
2. Security built in, not bolted on. Enterprise customers expect due diligence-ready security. Building it from the start is faster than retrofitting it.
3. AI as a product feature, not a marketing checkbox. Every AI component solved a specific user workflow problem — triage, drafting, scoring, qualification. No AI for AI’s sake.
4. Cloud-native from the start. Containerized services on Kubernetes gave us independent scaling, simple deployments, and infrastructure-as-code from week one.

Work With Me

If your company needs to move fast — whether it’s launching a new SaaS product, modernizing legacy infrastructure, or bringing experienced technical leadership to your team — I can help. I offer fractional CTO services tailored to your stage and goals, from strategic advisory to hands-on architecture and delivery.

Let’s talk about your project, or view my pricing to understand how engagements work.