About the role
As a RAG Engineer at TheWebVale, you will build retrieval-augmented generation systems that connect language models to private client knowledge bases. You will index documents, course materials, legal files, and medical records to generate grounded answers with verifiable source citations.
You will partner with backend engineers and product owners during our two-to-three week sprint cycles. You will tune chunking strategies, optimize hybrid search accuracy, and deliver grounded AI applications that clients rely on for daily work.
What you will do
- Design end-to-end RAG architectures connecting LLMs to proprietary client documents.
- Implement intelligent document parsing and chunking for PDFs, spreadsheets, and HTML.
- Configure vector databases including Pinecone, Qdrant, or PGVector for semantic retrieval.
- Implement hybrid search combining dense vector similarity with BM25 keyword matching.
- Integrate cross-encoder rerankers like Cohere to improve retrieved chunk relevancy.
- Measure retrieval quality using context recall, precision, and faithfulness benchmarks.
What you will need
- Hold 3–5 years of software development experience with deep specialization in RAG and search.
- Demonstrate advanced proficiency in Python, vector indexing, and embedding models.
- Show practical experience with vector stores like Pinecone, Qdrant, Weaviate, or PGVector.
- Understand the nuances of semantic search, chunk overlap, metadata filtering, and reranking.
- Possess strong skills writing automated evaluation tests with frameworks like Ragas or TruLens.
- Communicate retrieval trade-offs and accuracy metrics clearly to project stakeholders.
Nice to have
- Experience with advanced RAG patterns like parent-child chunking and hypothetical queries.
- Familiarity with unstructured document extraction tools like Unstructured or Marker.
- Knowledge of PostgreSQL full-text search integrated with vector similarity.