Sol — From Scattered Research to Shared Understanding

It acts as an always-on pulse of customer truth — centralising interviews, surveys, and field notes into a living system that's searchable, tagged, and queryable in plain English.

Anyone in an organisation can ask Sol a question — "What do customers find most frustrated about onboarding?" — and get instant, trustworthy answers drawn from real research, not assumptions.

This was never about building a commercial product; it was a design-led experiment in how AI can democratise research, accelerate discovery, and keep the customer's voice present in every decision.

Context

As teams scale, research knowledge often fragments across files, decks, and tools. Insights become hard to find, harder to trust, and almost impossible to reuse.

I wanted to see how far AI could go in solving that — not through another dashboard, but through a genuinely intelligent system that makes research accessible to everyone.

Action

Working solo in Cursor, I designed and built a multi-tenant SaaS platform from the ground up using Next.js 15, PostgreSQL, and a Fly.io-hosted MCP server for AI integration.

Key Features

• Workspace & Project structure for organisation-level data isolation
• AI querying via the Model Context Protocol (MCP) — allows ChatGPT and Claude to access workspace-specific data through secure API keys
• Role-based permissions (SuperAdmin → Viewer)
• Full-text search and tagging across Markdown, CSV, and text files
• Automated anonymisation to protect PII
• Custom domain and verified email setup (`heysol.io`, `noreply@heysol.io`)
• Professional front-end built with Tailwind, shadcn/ui, and Framer Motion
• Health monitoring, seed scripts, and test suites to ensure reliability

Deployed across Vercel (frontend) and Fly.io (AI server), Sol runs with full SSL, connection pooling, and role-based security — a functioning SaaS platform ready for enterprise use.

Impact

• Delivered a production-ready, AI-integrated research system with a 100% health-check pass rate
• Demonstrated that research data can become conversational, not static
• Validated a new model for AI-assisted ResearchOps, where AI supports discovery rather than replaces it
• Deepened my understanding of how design systems, data structures, and AI models intersect
• Reinforced the value of designing with AI — not just for it

Reflection

Building Sol taught me that the future of research isn't more documentation — it's living knowledge.

By bridging UX, data architecture, and AI integration, I've seen how design leaders can create systems that scale understanding, not just output.

Sol was my way of asking:

"What happens when everyone in an organisation can talk directly to their own research?"

The answer, I think, is a more informed, more empathetic product culture.

Stack & Deployment Summary

Layer	Tools
Frontend	Next.js 15, TypeScript, Tailwind, shadcn/ui, Framer Motion
Backend	Next.js API Routes, PostgreSQL / SQLite
AI Integration	MCP Server (Fly.io), Organisation-scoped API keys
Auth & Security	NextAuth.js, bcryptjs, RBAC, middleware protection
Hosting	Vercel (app), Fly.io (MCP), custom domain + SSL
DX & Testing	Health suite, seed scripts, TypeScript strict mode

What This Project Represents

"Sol is a design experiment brought to life in code.
It embodies how I think about AI, systems design, and the future of research: hands-on, responsible, and deeply human."