Browse categories
Explore
Fiverr Pro
English
$
USD
I will build a production rag chatbot over your documents
Waqar Makki
About this gig
Most RAG chatbots fail in production because they stop at chunk and embed. That works on 5 documents. It breaks at 500, on multi-page PDFs, and on any nuanced question.
I'm a production GenAI engineer based in Lahore. I've shipped RAG on AWS Bedrock (Llama 3 70B) for talent matching, and on OpenAI/Pinecone stacks for customer support. My systems are evaluated, not vibes-checked.
What you'll get:
Smart chunking tuned to your document structure not generic 512-token splits
Hybrid search (semantic + BM25 keyword) so exact terms still match
Metadata-rich embeddings + hierarchical indexes for long-document corpora
RAGAS evaluation report Faithfulness, Answer Relevancy, Context Precision & Recall
Source citations on every answer no hallucinations passed off as facts
Deployed demo, source code, README, 14-day post-delivery support
Stacks: AWS Bedrock (Llama 3, Claude), OpenAI, Anthropic, PGVector, Pinecone, ChromaDB, LangChain, LangGraph, FastAPI, Streamlit. I'll recommend what fits your budget and data volume.
Message me with a sample document and 5 expected questions I'll tell you honestly if it's a fit.
Get to know Waqar Makki
Waqar Makki
GenAI Specialist: LLMs, NLP, Computer Vision Expert
- FromPakistan
- Member sinceJul 2019
- Last delivery1 year
Languages
Urdu, English
I am a GenAI-focused Data Scientist & ML Engineer with over 4 years of experience specializing in production-grade NLP, GenAI, and Computer Vision applications. I translate complex R&D into high-impact commercial solutions.
Expertise:
- LLMs & RAG: Architecting AWS pipelines (Bedrock, PGVector) that reduced latency by 30%.
- Computer Vision: Expert in YOLOv8 and high-precision medical image segmentation.
- Agentic Workflows: Engineering autonomous AI ecosystems and REST APIs for rapid response.
I build scalable, optimized AI systems that deliver measurable results. Let’s collaborate!
FAQ
What document types do you support?
PDF, DOCX, HTML, Markdown, plain text, CSV, and websites (via crawl). Scanned PDFs need OCR — ask before ordering and I'll quote it as an add-on.
Do I need an OpenAI / AWS account?
Yes — the chatbot runs on your account and uses your API keys so you own the data and the bill. I'll guide you through setup.
How do you make sure it actually answers correctly?
I evaluate every system using RAGAS — Answer Relevancy, Faithfulness, Context Precision, and Context Recall. You get a report with the scores and the questions where it underperforms.
How much will the LLM API cost me to run?
It depends on traffic and document size. I'll size it before kickoff and recommend a model that fits your budget.
Can you deploy it for me?
Yes — Standard and Premium include deployment to AWS, Vercel, or your preferred platform with a public URL or API endpoint.
