Large Language Model Architect

Project Role : Large Language Model Architect

Project Role Description : Architect large language models (LLM) that can process and generate natural language. Design neural network parameters, trained on large quantities of unlabeled text data.

Must have skills : Machine Learning (ML)

Good to have skills : NA

Minimum 12 Year(s) Of Experience Is Required

Educational Qualification : 15 years full time education

Summary:

Own the end-to-end architecture of complex enterprise agentic AI solutions. Ensure systems are scalable, secure, observable, reliable, and integrated with enterprise data and applications.

Proven expertise selecting and governing LangGraph, OpenAI Agents SDK, Google ADK, Microsoft Foundry Agent Service, Vertex AI Agent Builder and Amazon Bedrock Agents, including MCP-based tool integration, identity, observability, evaluation and secure production deployment

Must have architected and delivered production AI or ML systems at enterprise scale. Reference architectures, demos, or vendor-platform configuration alone are insufficient.

Roles & Responsibilities:

• Design multi-agent and tool-using AI architectures for complex business workflows.
• Define orchestration, planning, state, memory, context, tool access, and human-approval patterns.
• Make architecture decisions across models, retrieval, applications, data, APIs, security, and cloud.
• Establish requirements for latency, throughput, resilience, cost, and auditability.
• Guide engineering teams from design through deployment and operations.
• Define evaluation, observability, guardrails, fallback, and incident-management patterns.
• Lead architecture reviews and senior client discussions.
  
  

Professional & Technical Skills:

• LLM orchestration, tool calling, workflow engines, RAG, and memory architectures.
• Distributed systems, APIs, event-driven architecture, cloud, and enterprise integration.
• LLMOps, tracing, evaluation, security, identity, access control, and cost optimization.
• Build-versus-buy and model-selection trade-offs.