Agentic AI Systems Architect / Developer

We are looking for an Agentic AI Systems Architect who can design and build production-grade AI systems beyond basic chatbots or simple RAG implementations. The ideal candidate should understand how to architect systems where AI agents can: • Plan and reason through multi-step tasks • Retrieve and validate information • Execute tool/function calls safely • Work with memory, state, and user context • Interact with APIs, databases, and internal systems • Operate in secure enterprise or government environments Basic knowledge of Node.js is required for API integration, backend connectivity, and service orchestration. Technical Requirements The candidate should have hands-on experience with: Agentic AI frameworks such as LangGraph, LangChain, CrewAI, AutoGen, Semantic Kernel, or custom orchestration layers Agent loops, planning/execution flows, task decomposition, reflection, validation, and human-in-the-loop workflows Tool calling / function calling, structured outputs, JSON schemas, API execution, and tool safety controls Stateful AI workflows, including state machines, graph-based orchestration, session state, and workflow persistence Advanced RAG pipelines, including: • Chunking strategies • Embedding model selection • Hybrid search • Metadata filtering • Query rewriting • Reranking • Context compression • Retrieval evaluation • Hallucination mitigation Memory architecture, including: • Short-term memory • Long-term memory • User-specific memory • Vector memory • Persistent knowledge stores LLM guardrails, including: • Prompt injection protection • Permission-aware retrieval • Output validation • Policy checks • Tool execution safety • Response verification LLM observability and evaluation, including: • Tracing • Prompt/version management • Eval datasets • Regression testing • Latency analysis • Cost monitoring • Failure analysis Enterprise integrations, including REST APIs, databases, CRMs/ERPs, document stores, webhooks, queues, and workflow engines Vector databases such as Qdrant, Weaviate, Pinecone, Milvus, Chroma, pgvector, Elasticsearch, or OpenSearch Secure deployment architectures for private cloud, on-premise, offline, or government-secure environments Node.js basics, including API development, async workflows, service integration, and connecting AI systems to backend applications Show more Show less

Agentic AI frameworks such as LangGraph, LangChain, CrewAI, AutoGen, Semantic Kernel, or custom orchestration layers; Agent loops, planning/execution flows, task decomposition, reflection, validation, and human-in-the-loop workflows; Tool calling / function calling, structured outputs, JSON schemas, API execution, and tool safety controls; Stateful AI workflows, including state machines, graph-based orchestration, session state, and workflow persistence; Advanced RAG pipelines, including chunking strategies, embedding model selection, hybrid search, metadata filtering, query rewriting, reranking, context compression, retrieval evaluation, hallucination mitigation; Memory architecture, including short-term memory, long-term memory, user-specific memory, vector memory, persistent knowledge stores; LLM guardrails, including prompt injection protection, permission-aware retrieval, output validation, policy checks, tool execution safety, response verification; LLM observability and evaluation, including tracing, prompt/version management, eval datasets, regression testing, latency analysis, cost monitoring, failure analysis; Enterprise integrations, including REST APIs, databases, CRMs/ERPs, document stores, webhooks, queues, and workflow engines; Vector databases such as Qdrant, Weaviate, Pinecone, Milvus, Chroma, pgvector, Elasticsearch, or OpenSearch; Secure deployment architectures for private cloud, on-premise, offline, or government-secure environments; Node.js basics, including API development, async workflows, service integration, and connecting AI systems to backend applications

Architect and build production-grade AI systems that enable agents to plan and reason across multi-step tasks, retrieve and validate information, execute tool calls safely, manage memory and user context, and interact with APIs, databases, and internal systems in secure enterprise environments.