Agentic AI Architect

Job description

About Position: We are seeking a highly experienced Agentic AI Architect with 10-15 years of industry experience to lead the design, architecture, and implementation of next-generation AI-driven solutions. The ideal candidate will possess deep expertise in Agentic AI frameworks, large language models (LLMs), autonomous AI systems, and cloud-native architectures. This role involves collaborating with business and technology stakeholders to define AI strategies, architect scalable intelligent agents, and deliver enterprise-grade AI solutions leveraging Vertex AI and modern cloud platforms. • Role: Agentic AI Architect • Location: Hyderabad & Bengaluru • Experience: 10 to 15 Years • Job Type: Full Time Employment What You'll Do: • Study the current business process, including users, decisions, handoffs, systems, information sources, exceptions, delays, controls and failure points. • Separate problems that need AI reasoning from those better solved through process correction, rules, search, analytics, workflow automation or conventional software. • Define the target process and show exactly where an AI assistant or agent participates, where deterministic controls apply and where a person must review or approve. • Convert business goals into measurable outcomes such as reduced handling time, improved first-time accuracy, fewer manual handoffs, faster information retrieval, lower rework or better control compliance. • Establish the current KPI baseline, target improvement, measurement method, data source, reporting frequency and business owner before claiming value. • Explain expected capabilities, limitations, uncertainty, failure modes, dependencies and residual risks in language understood by business and technology stakeholders. • Design end-to-end architectures covering models, RAG, knowledge sources, agent orchestration, tools, APIs, state, memory, user channels, identity, permissions and human escalation. • Define clear boundaries for agent autonomy, including which actions are read-only, which require confirmation and which must never be delegated to an agent. • Design for secure data access, tenant separation, privacy, audit evidence, prompt-injection resistance, least-privilege tools and controlled handling of sensitive information. • Define the evaluation strategy, including representative business scenarios, test datasets, acceptance thresholds, tool-action checks, safety tests and regression gates. • Define production requirements for response time, availability, throughput, cost, logging, tracing, quality monitoring, incident handling, rollback and model or prompt changes. • Select appropriate models, retrieval approaches and frameworks based on quality, security, latency, cost, portability and operating constraints rather than vendor preference. • Demonstrate deep proficiency in at least one AI development stack while adapting designs to enterprise-developed platforms, standards and custom frameworks. • Build or guide technical prototypes to test the highest-risk assumptions before committing to full implementation. • Review solution designs, code, prompts, tool contracts, data flows, test evidence and production-readiness material produced by engineering teams. • Create architecture diagrams, decision records, capability boundaries, risk assessments, cost models, delivery roadmaps and operating-model documentation. • Provide technical direction to AI engineers, QA engineers, data engineers, application engineers, security specialists and operations teams. • Track whether deployed solutions achieve the agreed KPIs and recommend tuning, redesign or retirement when the expected value is not being realized. Expertise You'll Bring: • 10+ years in software, data, cloud or solution architecture, including substantial hands-on experience with production AI or LLM-based systems. • Strong understanding of business-process analysis, value-stream mapping, decision points, control requirements and measurable outcome design. • Deep technical knowledge of LLMs, prompting, structured generation, embeddings, vector and hybrid search, reranking, RAG, tool use, agent orchestration, state and memory. • Demonstrable expertise in at least one production AI stack, such as Vertex AI and Gemini with ADK, LangChain/LangGraph, Semantic Kernel, LlamaIndex, or an equivalent platform. • Ability to review Python code, APIs, data pipelines, retrieval logic, agent workflows, automated tests and deployment designs at an engineering level. • Experience designing integrations with enterprise applications, APIs, data platforms, identity systems and approval workflows. • Strong understanding of AI evaluation, including answer quality, source support, task completion, tool-call correctness, safety, latency and cost. • Experience designing cloud-native, containerized or serverless solutions with CI/CD, secrets, IAM, observability and rollback controls. • Strong knowledge of responsible AI, privacy, secure AI architecture and controls for systems capable of taking actions. • Ability to create simple value models that distinguish projected benefit from measured benefit and avoid unsupported ROI claims. • Strong workshop facilitation, consulting, written communication and architecture-presentation skills. • Ability to challenge unsuitable AI use cases constructively and recommend a simpler solution where appropriate. • Production architecture experience with GCP services such as Vertex AI, Gemini, Vertex AI Search, BigQuery, Cloud Storage, Cloud Run, GKE, Pub/Sub, Cloud Logging, Cloud Monitoring and Cloud Trace. • Experience with knowledge graphs, document intelligence, multimodal systems, MCP or equivalent tool-integration protocols. • Experience designing AI solutions in regulated environments with strict audit, privacy, residency and approval requirements. • Familiarity with model and agent evaluation frameworks, experiment tracking and production AI monitoring. • Relevant cloud, mac