Vector Database & Embedding Engineer RAG Pipeline Development

Job Summary

Vector Database & Embedding Engineer

vector DBs (pgvector, Pinecone, Chroma, Milvus, Weaviate)

high-accuracy, high-recall retrieval systems

Key Responsibilities

1. Vector Database Design & Management

• Setup, configure and manage vector DBs such as:
  

• pgvector

  ,

  FAISS

  ,

  Pinecone

  ,

  Weaviate

  ,

  Chroma

  ,

  Milvus

• Design schemas for:
  
• Multi-embedding storage
  
• Metadata storage
  
• Document-level and chunk-level indexing
  
• Implement filtering, similarity search, MMR, reranking, and index optimization.
  

2. Embedding Pipeline Development

• Select, fine-tune, or run embedding models such as:
  
• Sentence-BERT, BGE, GTE, Instructor, FlagEmbedding
  
• OpenAI Embeddings / Azure OpenAI
  
• HuggingFace Transformers
  
• Build:
  
• Batch embedding pipelines
  
• Real-time embedding APIs
  
• Multi-encoder architecture for hybrid search
  
• Evaluate embedding quality, dimensionality, and vector drift.
  

3. Chunking, Indexing & Document Processing

• Design advanced

  chunking strategies

  :
• Fixed window chunking
  
• Sliding window
  
• Semantic chunking
  
• Layout-aware chunking (tables, lists, multi-column)
  
• Extract content from:
  
• PDFs, HTML pages, Office files, emails, scanned docs
  
• Build a complete indexing pipeline:
  
• Preprocessing Chunking Embedding Vector DB upsert Metadata linking
  

4. RAG Optimization & Retrieval Tuning

• Optimize retrieval for:
  
• Accuracy
  
• Latency
  
• Recall / diversity
  
• Implement hybrid search:
  

• Vector + Keyword

• Vector + Graph (GraphRAG)

• Build ranking stacks using rerankers (Cross-Encoders).
  

5. Backend & API Development

• Build APIs for:
  
• Document ingestion
  
• Embedding generation
  
• Retrieval & context merging
  
• Serve embedding + vector workflows using Python/FastAPI or Node.js.
  
• Integrate vector search with LLM prompt templates.
  

6. Monitoring, Evaluation & Scaling

• Evaluate retrieval metrics (precision@k, recall@k, MRR).
  
• Implement observability for indexing, failures, and accuracy degradation.
  
• Scale vector DBs horizontally & vertically based on dataset size.
  

7. Collaboration & Documentation

• Work with LLM engineers to design end-to-end RAG pipelines.
  
• Maintain documentation for:
  
• Embedding configs
  
• Chunking logic
  
• Vector schemas
  
• Retrieval settings
  
• Train internal teams on best practices.
  

Required Technical Skills

Vector Databases

• Strong hands-on with:
  

• pgvector

  (must-have for enterprise)

• Pinecone

  ,

  Chroma

  ,

  Weaviate

  ,

  Milvus

  , or

  FAISS

• Deep knowledge of:
  
• Index types (HNSW, IVFFlat, PQ, IVF-PQ)
  
• Similarity metrics (cosine, dot, euclidean)
  
• Index tuning (ef_search, ef_construction, cluster size)
  

Embeddings

• Experience generating and evaluating embeddings using:
  
• OpenAI / Azure OpenAI
  
• InstructorXL, BGE, GTE, FlagEmbedding
  
• Sentence-BERT / HF embeddings
  
• Knowledge of:
  
• Embedding dimensionality
  
• Tokenization & vector normalization
  
• Multi-embedding pipelines
  

Chunking & Preprocessing

• Strong experience with document processing libraries:
  
• PDFPlumber, PyMuPDF, Textract, Tika
  
• Designing chunking strategies for:
  
• PDFs
  
• Web pages
  
• Product catalogs
  
• Emails & logs
  
• Metadata creation and linking strategies.
  

Backend / Engineering

• Python (preferred), Node.js
  
• FastAPI / Flask
  
• SQL & NoSQL
  
• ETL pipelines (Airflow / custom)
  
• Docker, Linux environments
  

Experience Required

• Total Experience:

  26 years

• Relevant Vector Search / Embedding Experience:

  1–3 years
• Experience in building real RAG systems (highly preferred).
  

Preferred Skills

• Knowledge of:
  
• LangChain or LlamaIndex
  
• Rerankers (Cross-Encoders)
  
• Hybrid retrieval
  
• Graph + Vector hybrid search
  
• Experience in:
  
• OCR processing
  
• Data extraction
  
• Enterprise search systems
  
• Familiarity with:
  
• RedisSearch
  
• ElasticSearch vector search