AI Product Design
A user-owned AI assistant that acts across your tools.
Qilo turns natural language into controlled, multi-step actions across the tools professionals already use, running entirely on infrastructure the user controls. I owned it end to end, from concept and research to brand, system design, and the build.
Role
AI Systems Designer
Agent Workflow Builder
Timeline
10 weeks
Winter 2025
Stack
Figma, N8N, Pinecone, OpenAI, DigitalOcean, WordPress
The Problem
Knowledge workers are not blocked by thinking.
They are blocked by coordination.
Work lives across email, calendars, documents, and CRMs that do not talk to each other. The user becomes the integration layer between their own tools. The tools that could fix this asked for a shared cloud, which professionals with sensitive data will not accept.
The core tension: capability versus control.
The System
An orchestration engine, not a chatbot.
The entire system runs on a private server that belongs to the user.
All workflows, memory, and tool execution run inside user-owned infrastructure.
Only minimal context is sent to the LLM per request.
The Interface
A plugin that takes anything in.
The entry point, built as a WordPress plugin: text, image, file, or voice, bridged straight into the engine.
What I Built
Designed and built in parallel, not handed off.
I owned every layer, from concept, research, and brand through product and system design to the build itself.
Built
Multimodal orchestration backend
Built first
Production RAG and memory
Built on top
Agent and control model
Built
Front-end input layer
Failed or uncertain steps pause and return to the
user before anything downstream continues.
The Website
Built to remove doubt, in order.
Capability, proof, integrations, setup, privacy, then pricing as ownership.
Design System
Restraint as a strategy, not an aesthetic.
Restraint, white space, and a focused accent system reduce intimidation and reinforce clarity.
Key Decisions
Four decisions, and what each one cost.
Decision
User-owned infrastructure
A harder, more technical onboarding.
