AI assistant

Interactive assistant bot for informational support for parents of autistic children and autistic adults

Task

Build an information-support tool for a vulnerable audience — parents of autistic children and autistic adults — that relies only on verified content, helps users navigate topics and phrase requests, avoids overload, and behaves predictably and ethically. The assistant must act as a guide through a curated knowledge base, not a general-purpose chat buddy.

What I did

Designed and built the project end-to-end from scratch: concept, architecture, UX, copy, and logic.
Implemented strict AI behavior constraints: answers only from a curated knowledge base, no improvisation or “guessing”, no unnecessary advice.
Built interaction flows that reduce cognitive load: topic navigation, request clarification, calm UI states.
Implemented predictable session/state logic so users always understand context and avoid abrupt behavior changes.
Refined the tone: moderate, functional empathy without emotional pressure or imitation of “human counseling”.

Technical implementation

Modular architecture that separates core logic and makes maintenance and iteration easier.
Knowledge-source control: responses are generated only from a preloaded, verified knowledge base.
User registration and access management (statuses and access flows).
Payment integration and subscription management (activation, renewal, status).
Session and UI-state management for a predictable user experience.

Administration & scalability

Admin workflow focused on controllability: updating the knowledge base via file uploads into the database without touching the assistant’s core logic.
Clear separation between user-facing and admin logic so content updates don’t break assistant behavior.
Scalable by content and flows: the knowledge base and scenarios can grow without degrading quality.
Predictability as a product principle: stable behavior matters more than “wow effects”, especially in a sensitive domain.

Result

Summary

A predictable, safe assistant that reduces anxiety, helps users navigate the topic, and earns trust through control, clarity, and stable behavior. The project demonstrates a responsible approach to AI in a sensitive social domain: fewer effects — more quality, ethics, and controllability.

Project link

https://assistant.aineurosol.comProject temporarily paused by the client