Browse categories
Explore
Fiverr Pro
English
$
USD
I will build a custom rag chatbot trained on your documents and data
Shafay
About this gig
Your team already has the answers. They are just buried in PDFs, wikis and tickets. I build a chatbot that finds them and shows its sources.
I am a Lead Data and AI Engineer with 9 plus years of experience. I have built data and ML systems at Amazon, Delivery Hero, Goldman Sachs and Orion, and I hold 13 cloud certifications including AWS Machine Learning Specialty and AWS Solutions Architect.
WHAT I BUILD FOR YOU
Ingestion for PDFs, Word, Notion, Confluence, websites and databases
Chunking, embeddings and hybrid search with reranking for accurate retrieval
Answers with citations, and an honest I do not know instead of an invented one
An embeddable web widget, a Slack or Teams bot, or your own UI
Access control so private documents stay private
An evaluation set so you can measure accuracy instead of guessing
STACK
Python, LangChain, OpenAI, Claude, AWS Bedrock, pgvector, Pinecone, Qdrant, FastAPI, React
WHAT YOU GET
Documented source code you own, a working deployment and a handover call.
Send me a few real questions your team asks and I will tell you honestly whether RAG will answer them well.
Get to know Shafay
Shafay
Lead Data and AI Engineer, AWS Pipelines, LLM and RAG Apps
- FromGermany
- Member sinceAug 2026
Languages
English, Urdu, German
I build production data platforms and GenAI applications. 9+ years leading data and ML engineering at Orion, Delivery Hero, Amazon and Goldman Sachs.
What I do:
- Data pipelines and warehouses on AWS and GCP: Glue, Redshift, Snowflake, BigQuery, dbt, Airflow, Spark
- GenAI: RAG chatbots over your documents, AI agents, Bedrock, OpenAI, LangChain
- MLOps: SageMaker, Docker, Kubernetes, Terraform, CI/CD
13 cloud certifications including AWS Machine Learning Specialty, AWS Data Engineer, GCP and Terraform. Based in Berlin, working with clients across the EU and US.
My Portfolio
FAQ
What file types and data sources can you index?
PDFs, Word, PowerPoint, Markdown, HTML, Notion, Confluence, Google Drive, SharePoint, public websites and SQL databases. If a source has an API or an export, it can almost always be indexed. Send me a sample and I will confirm before you order.
How do you stop the chatbot from making things up?
Every answer is grounded in retrieved passages and shows the sources it used. When retrieval finds nothing relevant, the bot says so rather than guessing. I also build an evaluation set from your real questions so accuracy is measured, not assumed.
Where does my data live, and is it kept private?
Your choice. I can deploy entirely inside your own AWS or Azure account, or run a self hosted open model so nothing leaves your infrastructure. You keep your own API keys and pay providers directly, and I sign an NDA on request.
