Project Overview
The client, an executive coach, had proven frameworks and a working prototype as a custom GPT. But a custom GPT lives on someone else’s platform: no control over the experience, the data, the pricing, or the roadmap.
The business problem was ownership. To turn the method into a product, the coaching IP needed its own home, with the privacy promises a coaching relationship requires.
We designed and built the full platform: the AI product behind a login, the public site and waitlist, and the infrastructure that runs it all.
Key Challenges
- Coaching IP trapped in a custom GPT on a rented platform
- Conversations that must stay private and never train outside models
- Coaching, not therapy: the AI has to hold that line
- Real-time streaming chat that feels like a conversation
- A codebase the business can build on for years
AI products are easy to demo and hard to run. The job was to make the coach’s method reliable in production: fast, private, on brand, and fully under their control.
The Solution
We built the product as a modern web application: a Vue front end, an API with background workers for the AI calls, and responses that stream to the user in real time over a live connection.
The coach’s frameworks drive the AI’s behavior. Sessions are structured like real coaching work, and the system is deliberately scoped to clarity and decisions, not therapy.
- Product Design and Build
Two experiences in one product: a public site with a waitlist, and the coaching product itself behind an invite-only login.
- AI Engineering
LLM calls run through a queued worker and stream to the browser as they generate, so long thinking never blocks the app. The frameworks are encoded as the system’s foundation, not pasted into prompts by hand.
- Privacy by Design
Conversations stay in the client’s own database. Nothing is sold or used to train outside models, and the product says so plainly.
- Staging and Releases
A full staging environment mirrors production with isolated data. Every change ships through pipelines: merge, test, deploy, release.
- Owned Infrastructure
The client’s app, API, data stores, and domains all run on infrastructure they control, with documentation for whoever works on it next.
The Result
MindsetCoach AI is live: an invite-only product with a public waitlist at mindsetcoach.ai.
The coach went from renting a prototype to owning a product: the brand, the experience, the data, and the roadmap.
We continue to develop the product with the client as it grows toward wider release.