← Back to Projects

Swaad

Recommendation AI

Agentic RAG pipeline with self-reflective retrieval and multi-dimensional embedding space for preference modeling. Unstructured document ingestion via OCR with domain-specific entity extraction.

PythonRAGPineconeLangChainReactAgentic RAGVector SearchNLP

About

Agentic RAG system with self-reflective retrieval across multi-dimensional embedding space. OCR and entity extraction produce structured representations from unstructured inputs. Self-correcting agent behavior through reflection loops.

The Problem

Unstructured documents present information in inconsistent formats. Existing recommendation approaches use aggregate signals that cannot model individual preferences at granular levels.

The Approach

Multi-stage agentic pipeline. OCR and entity extraction produce structured representations. Custom multi-dimensional embedding space. Self-reflective agent evaluates candidate outputs through reflection loops.

Tech Stack

  • Frontend: React 18, TypeScript, TailwindCSS
  • Backend: Python 3.11, FastAPI, MongoDB
  • AI/ML: LangChain, Pinecone, OpenAI API, OCR Pipeline, Agentic RAG Framework

Apply

Apply by April 30, 2026

Slack: #swaad

Apply Now →

You'll learn

  • Agentic RAG
  • Vector Databases
  • Pinecone
  • Embedding Models
  • OCR Pipelines
  • Self-Reflective Agents

Roles

Rashanjot Kaur - AI Architect

Designed agentic RAG pipeline with self-correcting retrieval loops.

6 open role(s)

Applied AI Engineer - Open

Extend agentic RAG, reflection loops, OCR and embedding pipelines, and Pinecone retrieval quality.

Apply →

Backend Engineer - Open

FastAPI services, MongoDB, document ingestion, and orchestration for the agentic recommendation stack.

Apply →

Frontend Engineer - Open

Build recommendation interface, preference visualization, and clear exploration of multi-dimensional embeddings.

Apply →

Data Engineer - Open

OCR output normalization, entity extraction pipelines, and structured data for vector indexing.

Apply →

AI Researcher - Open

Benchmark retrieval and preference modeling; study embedding spaces, ranking, and self-reflective agent behavior.

Apply →

DevOps / MLOps - Open

Deploy FastAPI/React stack; monitor APIs, Pinecone usage, and pipeline reliability.

Apply →