Ai Aws Architect Job at Realign (Parsippany)

Job Responsibility

Architect and design an end-to-end solution for an AI-driven Retrieval-Augmented system
Define and implement the embedding strategy, document chunking methodology, and ranking techniques to ensure high-quality retrieval
Evaluate, select, and integrate appropriate Large Language Models (LLMs) using Amazon Bedrock
Design and implement a scalable vector database solution (such as OpenSearch, Qdrant, or MongoDB Atlas with vector search)
Architect and deploy the cloud infrastructure leveraging AWS services including Amazon Bedrock (LLM), SageMaker, AWS Lambda, and Amazon EKS
Establish MLOps best practices by designing CI/CD pipelines for model deployment and managing model registry updates
Apply FinOps principles to balance cost and performance, optimizing model size, inference latency, and caching strategies
Design secure access controls and implement secrets management using AWS Secrets Manager

10+ years of experience in Software Development & ML engineering, including 4+ years in Solution architecture or AI architecture delivering production-grade systems
Strong hands-on expertise in Python (workflow orchestration, model evaluation), Node.js, and experience working with agent-based frameworks such as AutoGen, LangChain, or ADK. Solid knowledge of prompt engineering techniques and tool-calling mechanisms
Deep understanding of retrieval-augmented systems, including embeddings, chunking strategies, ranking mechanisms, and Vector databases such as Qdrant and MongoDB Atlas with vector search capabilities
Extensive experience in MLOps practices, including CI/CD pipelines for ML models, model registry management, and containerized deployments using Kubernetes
Strong cloud expertise in AWS, with hands-on experience in services such as Amazon Bedrock, SageMaker, AWS Lambda, and Amazon EKS
Demonstrated experience in implementing Responsible AI practices, security best practices (encryption, secrets management, network isolation), and FinOps strategies for AI workloads, including cost and latency optimization