Back to system map KK
case_study 01 · agentic expert system

AI Consulting System

LLM & AgentsGemini 2.0 Flash2024 · Live

A production expert system where one interface routes users to four specialized LLM consultants — Education, Healthcare, Finance and Retail — each with its own domain-tuned prompting strategy. The architecture became the basis of my published research paper, "AI-Driven Consulting: A Smart Expert System for Diverse Domains."

AI Consulting System screenshot

01 What problem does it solve?

A single general-purpose chatbot performs poorly across specialized consulting domains: answers drift, tone is generic, and domain constraints (medical caution, financial disclaimers, retail context) get lost. The goal was one product that behaves like four different domain experts without four separate apps.

02 How does it work?

user querydomain routerdomain-tuned prompt chaingemini 2.0 flashexpert response

03 What did it achieve?

4expert domains
1published research paper
prodlive deployment

The system ships in production and the architecture was formalized into an international research publication. It demonstrates the pattern I use at work daily: specialized LLM agents beat one generic model when each carries domain-scoped prompting and constraints.

04 Stack

Gemini 2.0 FlashPrompt EngineeringNext.js 15TypeScriptFirebaseClerkRadix UITailwind CSS
View source Launch live demo