Browse categories
Explore
Fiverr Pro
English
$
USD
I will build a rag ai chatbot that chats with your pdfs and documents
Marek Kulma
About this gig
Stop your team from digging through PDFs. Let them just ask.
I build custom AI chatbots that answer questions directly from YOUR documents contracts, manuals, policies, reports, FAQs with the exact source passage shown for every answer. No hallucination, fully verifiable.
This is NOT a ChatGPT wrapper. Generic bots invent answers about your documents. Mine uses real RAG (Retrieval-Augmented Generation) with semantic search: it finds the relevant passage first, then answers only from it. Ask "how much vacation?" and it finds your "annual leave" clause even though the words differ. Meaning-based search, not keyword matching.
What you get:
Semantic search handles synonyms & rephrased questions
Chat with any document PDFs, manuals, contracts, FAQs
Cited answers every response shows its source
Private embeddings run locally
Fast high-speed LLM inference
Perfect for: customer support, HR, legal, internal knowledge bases, onboarding.
You get the full source code, deployed on YOUR account you own it, never locked in.
Message me before ordering with your use case and document count I'll confirm the right package.
Get to know Marek Kulma
Marek Kulma
Automation and Computer Vision Developer
- FromPoland
- Member sinceMay 2026
- Avg. response time1 hour
Languages
Polish, English, German
I build automation and data-extraction systems that save businesses
hours of manual work. My focus is turning messy real-world documents —
invoices, financial reports, scraped data — into clean, structured
output you can actually use.
I work in Python with custom-trained extraction models (YOLOv8, OCR),
plus automation tools like n8n and Make.com. Whether it's a one-off
batch of PDFs or a recurring workflow, I deliver production-ready
solutions, not throwaway scripts.
Send me a sample of what you're working with and I'll tell you straight
whether I can help.
FAQ
Is this just ChatGPT with my documents pasted in?
No. This is a proper RAG system — it retrieves the relevant part of your document by meaning and answers only from that, so it doesn't hallucinate and shows you the exact source. It works reliably on large documents where pasting everything into ChatGPT fails.
Will my documents stay private?
The semantic search runs locally, so your document text isn't sent to a third-party embedding service. The final answer step uses an LLM API (Groq). If you need a fully offline setup, message me — that's available as a custom order.
Do I own the chatbot, or am I locked in?
You own it completely. I deliver the full source code and deploy it on your own account. You're never dependent on me to keep it running.
What documents work?
PDFs, Word docs, and text files. Scanned or image-based PDFs need OCR, which is included in the Premium package (or available as an add-on).
Are there running costs?
The chatbot calls an LLM provider (Groq) for each answer. Groq has a generous free tier that covers light-to-moderate use. For heavy use I'll set you up with the cheapest paid option and show you the costs upfront.
I'm not sure which package I need.
Just message me before ordering with your use case and roughly how many documents or pages you have. I'll point you to the right package honestly — no upselling.
