LiveIT & Software Specialist (Dentech Medikal)

AI Customer Support Chatbot

Domain-Specific Knowledge Base Chatbot & Support Resolution System (chatbot_v1)

JavaScriptNode.jsOpenAI APIPrompt EngineeringREST APIs

Problem

Support teams spent excessive hours answering over 100 domain-specific business queries manually based on internal document files, creating support bottlenecks.

Solution

Engineered an AI-powered customer support chatbot (chatbot_v1) integrated with internal knowledge bases (20 documents), successfully automating responses for domain-specific business queries with prompt engineering & guardrails.

System Architecture

Node.js JavaScript server serving knowledge base text chunks to OpenAI GPT API with custom prompt guardrails and fallback escalation.

System Architecture Flow

3 Components

STEP 01

Chat Interface

Embedded customer support chat widget.

STEP 02

Chatbot Engine (chatbot_v1)

Prompt engineering & guardrail orchestrator.

STEP 03

Knowledge Base Store

Internal business documents repository (20 docs).

Execution Workflow

1. Customer submits query on web chat.

2. Prompt context builder injects relevant knowledge base policy info.

3. LLM validates output against guardrails and formats clear response.

4. Complex out-of-scope issues automatically routed to human specialist.

Business Impact & Metrics

Significantly reduced manual support workload for Dentech Medikal, providing instant responses 24/7.

Automated Domain Queries100+

Before: 0% automatedAfter: 100+ queries

Knowledge Docs Integrated20 Docs

Before: Manual PDF lookupAfter: Instant retrieval

Key Lessons Learned

Strict prompt guardrails prevent the AI model from inventing non-existent company policies or prices.

What I Would Build Differently Today

“I would use LangGraph to manage conversational state cycles, ensuring fallback routing triggers cleanly when confidence scores are low.”