AI Engineer @ W2

AI Engineer

Charlotte, NC or Dallas, TX

Long Term

Contract

AI System Requirements: Embedding Models + Knowledge Graphs + Advanced RAG + Ontology Extraction

Objectives

- Enable semantic search and reasoning over domain-specific data.

- Integrate embeddings with knowledge graphs for hybrid retrieval.

- Support advanced RAG pipelines for contextualized generation.

- Automate ontology extraction to enrich structured knowledge bases.

Core Components

- Python Frameworks: PyTorch, TensorFlow, HuggingFace Transformers.

- Embedding Models: Sentence-BERT, OpenAI embeddings, domain-specific fine-tuned models.

- Knowledge Graphs: Neo4j, RDF/SPARQL, graph-based reasoning engines.

- Advanced RAG: Hybrid retrievers (vector + symbolic), context re-ranking, multi-hop reasoning.

- Ontology Extraction: NLP pipelines for entity/relation extraction, schema induction, ontology alignment.

. Functional Requirements

- Data ingestion and preprocessing (structured + unstructured).

- Embedding generation and storage in vector DB (e.g., Pinecone, FAISS, Weaviate).

- Knowledge graph construction and querying.

- Retrieval pipeline combining embeddings + KG queries.

- Ontology extraction from text corpora to update KG schema.

- RAG pipeline for contextualized text generation with grounding.
. Non-Functional Requirements

- Scalability: handle millions of documents and graph nodes.

- Accuracy: embeddings must achieve 90% semantic similarity on benchmark tasks.

- Latency: retrieval + generation under 2 seconds per query.

- Extensibility: modular design for plugging in new models or KG schemas.

Integration Architecture

- Frontend: API endpoints for query and generation.

- Backend: Python services orchestrating embeddings, KG queries, and RAG.

- Storage: Vector DB + Graph DB.

- Pipeline: Ontology extraction KG enrichment Hybrid retrieval RAG generation.

Munesh

,

CYBER SPHERE LLC