I will build a custom ai chatbot with rag
About this gig
Turn your business knowledge into a custom AI chatbot.
I led AI development for Tax Canvas, an AI platform for tax professionals, building RAG workflows, OCR document processing and source-backed answers with LangChain, LangGraph and vector databases.
I bring that hands-on experience to chatbots for customer support, internal knowledge search and business websites.
Depending on your package, I can provide:
- RAG over documents, website content or business data
- Source citations so users can check answers
- OCR for scanned documents
- A custom chat interface and deployment
- User login, saved chat history and database integration
- Source code and setup instructions
I'm offering introductory rates for my new RAG service on Fiverr.
Each package has defined data and feature limits. API, hosting and third-party usage fees are paid separately by you. Complex integrations and agent workflows require a custom quote.
Message me with your use case, sample data and required integrations so we can confirm the right scope before you order.
Get to know Taewon Sohn
- FromSouth Korea
- Member sinceNov 2020
- Last delivery2 years
Languages
English, Korean
FAQ
1. What do you need to get started?
Please share your use case, sample documents or website URLs, preferred language and any required integrations. For deployment, you will also need your own hosting and AI provider accounts.
2. Are API and hosting costs included?
No. Package prices cover development. LLM, embedding, OCR, hosting and other third-party usage fees are separate and paid through your own accounts.
3. What data limits apply to each package?
Basic covers 1 source: up to 50 text-based pages or 10 static web pages. Standard and Premium cover up to 3 sources and 150 pages total, including up to 30 scanned pages for OCR. Larger or complex datasets require a custom quote.
4. Will the chatbot always give correct answers?
No AI system can guarantee perfect accuracy. I use retrieval, source citations and instructions to acknowledge missing information to reduce unsupported answers. Results depend on your data and are checked against agreed sample questions.
5. What counts as a revision?
A revision covers adjustments within the agreed scope, such as prompts, retrieval settings or interface styling. New features, integrations or additional data are quoted separately. Fixing defects in agreed functionality does not require a paid extra.

