Case Study · AI Products — SoftsensorX

Conversational RAG over a Global Knowledge Library

Spiritual & Wellbeing Foundation · RAG Pipeline · Semantic Search · Conversational AI · Source Grounding

The challenge

Decades of teaching, hard to reach.

A vast, unsearchable corpus

Thousands of talks, articles and Q&A sessions accumulated over decades — rich, but impossible to navigate by keyword or menu.

Seekers want answers, not links

People arrive with a question in their own words and need a direct, contextual answer — not a list of hundreds of videos to sift through.

Fidelity to the source is non-negotiable

Answers had to reflect the foundation's own material faithfully — a generic LLM inventing content was unacceptable.

Global scale and many topics

A worldwide audience and a wide range of subjects meant retrieval had to stay precise across a large, diverse knowledge base.

What we built

A grounded conversational assistant.

Ingestion & semantic indexing

The content library is chunked, embedded with sentence-transformers and indexed in a Pinecone vector store with metadata — organized into namespaces for clean, precise retrieval.

RAG-Fusion retrieval

Multi-query RAG-Fusion with hybrid semantic + keyword search and re-ranking surfaces the most relevant passages for each question, across a large and varied corpus.

Grounded, streaming answers

Responses are generated strictly from retrieved passages and streamed over WebSockets with conversation memory — a natural, real-time chat experience.

Source-linked & on-message

Every answer stays faithful to the foundation's own material and can point back to the sources it drew from — trustworthy by construction.

Outcome

What good looks like.

Conversational

Natural-language access to decades of talks and teachings.

Grounded

Answers drawn strictly from the foundation's own corpus.

Corpus-scale

Precise retrieval across a large, multi-topic knowledge base.

Foundational

An early SoftsensorX RAG platform — reused across finance, ESG and documents.

Explore NLP, LLM & RAG →  ·  ← All case studies  ·  Talk to our team →