What are the responsibilities and job description for the Onsite--Sr Gen AI Engineer (GenAI + Agentic Systems + RAG) position at Cliff Services Inc?
Greetings,
Role: Title: Principal Gen AI Scientist (Hands-On GenAI Agentic Systems RAG Architect)
Location: on-site M-F in Mclean VA
Duration: Long Term
Local to Virginia preferred
Must Have Qualifications: Must have hands on experience with machine learning transitioned into GenAI. Rag, Python- Jupyter, other Software knowledge, using agents in workflows, strong understanding of data. Preferred: Built AI agent, MCP, A2A, Graph Rag, deployed Gen AI applications to production.
**Key Responsibilities: **
* Architect and implement scalable AI Agents, Agentic Workflows and GenAI applications to address diverse and complex business use cases.
* Develop, fine-tune, and optimize lightweight LLMs; lead the evaluation and adaptation of models such as Claude (Anthropic), Azure OpenAI, and open-source alternatives.
* Design and deploy Retrieval-Augmented Generation (RAG) and Graph RAG systems using vector databases and knowledge bases.
* Curate enterprise data using connectors integrated with AWS Bedrock's Knowledge Base/Elastic
* Implement solutions leveraging MCP (Model Context Protocol) and A2A (Agent-to-Agent) communication.
* Build and maintain Jupyter-based notebooks using platforms like SageMaker and MLFlow/Kubeflow on Kubernetes (EKS).
* Collaborate with cross-functional teams of UI and microservice engineers, designers, and data engineers to build full-stack Gen AI experiences.
* Integrate GenAI solutions with enterprise platforms via API-based methods and GenAI standardized patterns.
* Establish and enforce validation procedures with Evaluation Frameworks, bias mitigation, safety protocols, and guardrails for production-ready deployment.
* Design & build robust ingestion pipelines that extract, chunk, enrich, and anonymize data from PDFs, video, and audio sources for use in LLM-powered workflows—leveraging best practices like semantic chunking and privacy controls
* Orchestrate multimodal pipelines** using scalable frameworks (e.g., Apache Spark, PySpark) for automated ETL/ELT workflows appropriate for unstructured media
* Implement embeddings drives—map media content to vector representations using embedding models, and integrate with vector stores (AWS KnowledgeBase/Elastic/Mongo Atlas) to support RAG architectures.
Thanks & regards,
K Hemanth | Recruitment Specialist
Email: hemanth.k@cliff-services.com