Browse categories
Explore
Fiverr Pro
English
$
USD
I will build a rag chatbot for your website trained on your own documents
Ohidur
About this gig
A generic AI chatbot makes things up. A RAG chatbot quotes your actual documents.
I build retrieval-augmented (RAG) chatbots that answer from YOUR manuals, policies, contracts, product docs or support history, and show which page the answer came from.
WHAT YOU GET
An AI chatbot trained on the documents you send
Cited answers, so your team can verify every claim
An embeddable chatbot widget for your website, or a standalone chat page
An honest accuracy report: I test it on real questions and show you where it is weak
Source code and a re-indexing script you own
TWO WAYS TO RUN IT
Cloud: fastest and cheapest, on the OpenAI or Claude API.
Private: the whole RAG stack runs on YOUR server with an open model. Nothing goes to a third party. Choose this for client records, medical, legal or HR documents.
WHY ME
Five years of production software and a robotics job that runs on tests, not vibes. I will tell you before you order if your documents are too thin for RAG to work well, rather than shipping a chatbot that confidently guesses.
Send a sample document and your three hardest questions. I will tell you what accuracy to expect.
Get to know Ohidur
Ohidur
AI and Automation Engineer, MCP, RAG, Claude Training
- FromBangladesh
- Member sinceJan 2020
- Last delivery2 years
Languages
English, Bengali
Robotics software engineer. I build the software that makes machines and AI systems actually do things.
Day job: the full software stack for a collaborative-robot cell - digital twin, controller RPC client, Isaac Sim, and PLCs read over Modbus and RS-485.
Here: MCP servers that connect Claude to your tools, RAG chatbots trained on your documents, ESP32 and PLC dashboards, Docker deployment on your VPS, and live 1-on-1 Claude training.
Five years shipping production software and 200+ freelance projects delivered. I test what I ship, I document it, and I say when an idea will not work.
FAQ
What file types can you index?
PDF, Word, Markdown, HTML, plain text, CSV, Notion exports, Google Docs and Sheets, and public web pages. Scanned PDFs need OCR first, which I can add - tell me and I will quote it.
Will it make things up?
Far less than a plain chatbot, because it answers from retrieved passages and cites them. I also configure it to say "I could not find that in your documents" rather than inventing an answer, and I show you the accuracy test before delivery.
Does my data get sent to OpenAI?
In Basic and Standard, yes - your document chunks go to the model API. In Premium, no: the model, the vector database and the app all run on your own server. Pick Premium if the documents are confidential.
What does it cost to run each month?
Cloud: usually a few dollars a month in API calls for a small internal bot, more with heavy public traffic. Self-hosted: just your server, roughly $20-$80 a month for a machine that can run a good open model. I will size it for you.
Can I add documents later without paying you again?
Yes. I hand over a re-indexing script, and Standard and Premium include instructions for it. If you want a click-and-upload admin panel instead, message me and I will quote it.
How many documents is too many?
Packages cover up to 5000 pages. Beyond that it is still very doable but needs a different retrieval design - message me and I will quote a custom offer.
