Browse categories
Explore
Fiverr Pro
English
$
USD
I will build an ai chatbot that answers from your documents
Hasham Vakani
Level 2
About this gig
Give your customers and team instant answers from your own documents with real citations, not guesses.
I build AI chatbots that read your PDFs, docs, and knowledge base, then answer questions using only your content and link back to the exact source. No hallucinations, no made-up answers: when the answer isn't in your documents, the bot says so instead of inventing one.
I've shipped this in production (Claude + pgvector, live on the App Store and Play Store). What you get:
- A chatbot trained on your documents (PDF, DOCX, TXT, Markdown, web pages)
- Answers with citations pointing to the exact source page
- Honest "I don't know" when the answer isn't in your content
- An admin panel to add or update documents yourself (Standard and Premium)
- Your choice of a live website widget or an internal team tool
Great for customer support, internal knowledge bases, onboarding, and product docs.
Optional monthly plan: I keep your knowledge base fresh, re-index new documents, and tune answer quality as your content grows.
Before ordering, message me what documents you want it to answer from and roughly how many. I'll recommend the right tier
Get to know Hasham Vakani
Hasham Vakani
AI agents, RAG, automation and MCP, production builds not demos
Level 2
- FromPakistan
- Member sinceAug 2020
- Avg. response time1 hour
- Last delivery1 week
Languages
English, Spanish, French
I ship production AI for startups - agents, RAG, automation, and MCP servers - and I fix AI-built apps that broke in production.
8,700+ hours delivered. 100% satisfaction. Claude Code daily on live work.
What I build:
- AI agents (tools, memory, Claude/OpenAI)
- RAG chatbots that answer from your docs with citations
- AI automation with n8n or Make
- AI voice agents for inbound calls and bookings
- MCP servers and AI SaaS MVPs (Next.js, Stripe)
- Fixing broken AI apps (Lovable, Bolt, Cursor)
Tech: Next.js, TypeScript, Claude, OpenAI, AWS.
Message me your use case.
My Portfolio
Other AI Development Services I Offer
FAQ
What's the difference between this and ChatGPT?
ChatGPT doesn't know your documents. RAG retrieves YOUR data first, then asks the LLM to answer using only that retrieved context — with citations.
How accurate are the citations?
Each response includes the source document + page/chunk. Premium tier adds a re-ranking layer that improves citation hit-rate measurably (we benchmark it before delivery).
Which vector database?
Pinecone (managed, fastest) for production. pgvector (self-hosted, $0 ops cost) for smaller deployments. I'll recommend based on your scale.
Can I add new documents after delivery?
Yes. Standard and Premium include an admin panel to upload new docs without my involvement.
Will it hallucinate?
Significantly less than a non-RAG bot, especially with the re-ranking layer in Standard/Premium. We tune the system prompt to refuse when retrieval confidence is low.
Do you handle deployment?
Basic: code only. Standard: deployment guide. Premium: full deployment to your cloud.
What document formats are supported?
PDF, DOCX, TXT, Markdown, HTML out of the box. Premium can add Notion, Confluence, Google Drive, custom DB sources.
How long does it take to index 1,000 documents?
~30-90 minutes depending on doc length and embedding model. One-time cost, then incremental updates are fast.
What are the ongoing costs after delivery?
Two costs: (1) Vector DB — Pinecone free tier handles ~100K vectors, then $70/mo for paid. pgvector self-hosted is $0 ops cost. (2) LLM API — Claude or OpenAI charges per query, typically $20-100/mo for moderate usage. I'll size your monthly cost during kickoff so there are no surprises.
Do you sign NDAs?
Yes — always available before code or document access. Especially common for healthcare, legal, and fintech clients.
