Door categorieën bladeren
Ontdekken
Fiverr Pro
Nederlands
$
USD
I will build a custom rag chatbot with langchain,fastapi and openai
Arnav R
Over deze dienst
Do you need an AI chatbot that answers from YOUR data not made-up answers?
I build production-grade RAG (Retrieval-Augmented Generation) chatbots that:
Answer from your documents (PDFs, JSON, text files)
Cite sources for every answer
Say "I don't know" instead of hallucinating
Deploy to production (Render + Vercel
WHAT I'VE BUILT
A live RAG system for Indian college admissions:
shiksha-chat.vercel.app
It answers questions about 30+ IIT colleges and Bihar state institutions with cited sources, and achieved 100% accuracy on a 40-question evaluation (30 factual + 10 anti-hallucination tests).
MY TECH STACK
LangChain (RAG pipeline)
OpenAI (GPT-4o-mini + embeddings)
ChromaDB (vector store)
FastAPI (backend)
Next.js + Tailwind (frontend)
Render + Vercel (deployment)
WHAT YOU GET
Basic: Working RAG chatbot over your docs
Standard: RAG + custom Next.js chat UI
Premium: Full deployment + evaluation suite + 30-day support
WHY ME
I build production systems, not demos
Every answer is cited from your sources
I test for hallucinations before delivery
Full source code always included
Send me a message with your use case I'll reply within a few hours with a plan.
Maak kennis met Arnav R
Arnav R
Full Stack Developer
- Afkomstig uitIndia
- Lid sindsapr 2025
Talen
Engels
I am a Full Stack Developer who builds production-ready web applications. I built CourseAdmission, an AI-powered platform connecting students to 4,760+ colleges, featuring a custom-built AI chatbot and a full Next.js tech stack. I have experience in AI training, code quality, and debugging.
Veelgestelde vragen
What kind of documents can you use?
PDFs, text files, JSON, CSV, and markdown. If your data is in a database or a website, we can discuss extracting it — usually possible with a small custom script.
How is this different from a ChatGPT wrapper?
A ChatGPT wrapper just forwards your question to OpenAI. A RAG system first searches YOUR documents, then feeds the relevant parts to the AI. The result: answers grounded in your data with source citations, and the bot says "I don't know" when the answer isn't in your documents.
Will the bot hallucinate (make things up)?
No. The prompt is engineered to refuse answering when the context is missing. I also test with adversarial questions (things the data can't answer) to verify it doesn't guess.
Do I need to provide my own OpenAI API key?
Yes. You'll need an OpenAI API key (about $5 credit is enough for months of testing). I'll guide you on setting it up in the deployment step.
