GenAI Developer LLMs, RAG & Knowledge Graphs (Neo4j)

GenAI Developer LLMs, RAG & Knowledge Graphs (Neo4j)

Experience:

Role Overview

We re looking for a hands-on GenAI Developer to design, build, and productionize LLM-powered applications retrieval-augmented generation (RAG), conversational agents, document understanding, and decision support. You ll combine AWS Bedrock/SageMaker with open-source stacks and integrate a Neo4j knowledge graph to deliver reliable, explainable AI in a product environment.

What You ll Do

• Ship end-to-end GenAI features:

  data ingestion chunking/embedding RAG pipelines prompt orchestration evals deployment monitoring.

• Build robust RAG systems:

  hybrid retrieval (vector + keyword), query re-writing, reranking, citations, and grounding on enterprise data.

• Model orchestration on AWS:

  Bedrock model selection (Claude, Llama, etc.), prompt flows, guardrails, and cost/perf tuning; SageMaker for custom training/fine-tuning (LoRA/QLoRA/PEFT).

• Knowledge Graph (Neo4j):

  design schema/ontologies, build entity/relation extractors, create KG-augmented RAG, and run graph reasoning with Cypher & GDS.

• Eval & quality:

  define automatic evals (faithfulness, answer relevance, toxicity, latency), A/B tests, and red-teaming; maintain prompt and dataset versioning.

• Ops & reliability:

  implement LLMOps (observability, tracing, guardrails, rate/cost controls), CI/CD, IaC, and runtime monitoring.

• APIs & apps:

  expose services via REST/GraphQL; build lightweight UIs or chat backends, integrate auth, and log interactions for continuous improvement.

• Security & compliance:

  PII handling, prompt injection defense, output filtering, and enterprise IAM hygiene.

Core Skills & Tech Stack

Languages & Frameworks

• Python (FastAPI), TypeScript/Node.js (nice-to-have), SQL
• GenAI frameworks: LangChain / LlamaIndex / Haystack
• Prompt tooling & observability: Langfuse, Weights & Biases (or equivalents)

LLMs & Modeling

• Retrieval & embeddings: Bedrock embeddings, sentence-transformers, Cohere/HF embeddings
• Fine-tuning/distillation: LoRA/QLoRA/PEFT, bitsandbytes, Hugging Face ecosystem
• Rerankers & rewriters: cross-encoders, T5/FLAN or similar
• Safety/guardrails: content filters, groundedness checks, regex/policy engines

AWS (Focus)

• Bedrock

  (model access, guardrails),

  SageMaker

  (training, endpoints),

  ECR/ECS/EKS

  ,

  Lambda

  ,

  API Gateway

  ,

  S3

  ,

  CloudWatch

  ,

  Step Functions

• OpenSearch / Kendra

  for retrieval;

  RDS/ Aurora / DynamoDB

  for metadata/state

• IAM

  ,

  Secrets Manager

  ,

  VPC

  for security & networking

Knowledge Graphs

• Neo4j

  (schema/ontology design, Cypher, APOC),

  Graph Data Science (GDS)

• Entity/relation extraction pipelines; KG-RAG patterns; graph-based reasoning and path explanations

Vector/Search

• OpenSearch vector, pgvector, Milvus, or Pinecone (any 1 2 strongly)

MLOps / LLMOps

• MLflow or SageMaker Model Registry; Docker; CI/CD (GitHub Actions/GitLab)
• Tracing/telemetry, dataset & prompt versioning, cost/latency dashboards

Qualifications

• Bachelor s/Master s in CS/EE/Math or equivalent practical experience
• 5 8 years building data/ML systems; 2+ years specifically with LLMs/GenAI
• Proven production deployments on AWS and experience with Neo4j in real use cases