Rajiv U.Private profile: Upwork shows its earnings only to signed-in members, so we check it every few weeks, not daily. MRR here is a rough estimate (Upwork's six-month earnings ÷ months active). Last read Oct 9, 2026.

I build document Q&A and RAG systems that show where answers come from. My paid AI work includes a Bedrock agent for an insurance platform processing over 400,000 transactions a month, and a customer-support assistant that answers from company knowledge. Is your assistant missing useful information or giving unsupported answers? Send me two examples. If you're adding document Q&A to an existing app, tell me what your users need to find. I'll suggest a focused first milestone. I focus on two kinds of engagement. • Improve an existing RAG system. Review ingestion, chunking, metadata, embeddings, vector retrieval and answer generation. Agree a small set of test questions and supporting sources, then compare the system before and after the changes. Include questions it should decline to answer when the evidence is missing. • Add document Q&A to your application. Connect your documents to a Python/FastAPI service and your existing interface, with source references, document updates and access controls scoped to your needs. I work with OpenAI, Claude and Amazon Bedrock, using the stack that fits your application. My paid client work includes the following projects. • Insurance operations agent. Built an Amazon Bedrock agent for an insurance platform processing more than 400,000 transactions per month. I developed Python Lambda tools to query operational records through Athena, Glue and S3 and investigate transaction failures. My work included query integration, timestamp and type handling in generated SQL, and inference throughput. The client handled the production frontend and deployment. • Customer-support RAG assistant. Built a WhatsApp support agent using n8n, Claude and a company knowledge base. It answered routine questions about products, orders and policies, handed complex queries to a human with context, and logged conversations in Google Sheets. The client could add FAQs without a developer. • AWS integration recovery and maintenance. Recovered a live subscription integration where cancellation events were being dropped. I traced API Gateway, Go Lambda, DynamoDB and provider API requests, corrected the failures, and added safe replay, idempotency and bounded retries. I continue maintaining the system. You can also review two RAG portfolio demonstrations. • FastAPI document Q&A. A Google Drive knowledge base connected to a custom GPT through a Python retrieval API, using Supabase/PostgreSQL and pgvector. The demo returns source quotations and shows a newly added transcript becoming retrievable. • React and Bedrock assistant. A healthcare demo using synthetic data, with a React interface, Python Lambda, document retrieval and Athena Text-to-SQL. It shows document answers and structured-data queries within one application. For document AI running on AWS, I can also review the surrounding integration, permissions and failure handling. My paid infrastructure work includes a Terraform and Docker review covering 13 files and roughly 1,250 lines, with written findings on security, cost and deployment risks. We can begin with a focused paid assessment or one implementation milestone. Before starting, we'll agree the data sources, deliverables and how we'll judge whether the result is useful. You'll receive the agreed implementation or findings, test results and handover notes. I'm a Top Rated freelancer based in Amsterdam, with 18+ years in enterprise technology across AWS and Oracle. I currently work at AWS and take on independent freelance projects for around 20 hours per week. Send me two questions your system needs to answer, the documents or data it should use, and what's working today. I'll suggest a first step we can scope and price together. Core skills RAG, document Q&A, semantic search, vector search, Python, FastAPI, OpenAI API, Claude, Amazon Bedrock, PostgreSQL, pgvector, Supabase, React, AWS Lambda, API integration, n8n.