AI Solution Architect

Job Description

Information Technology (IT) acts as an internal function the developing, running, and maintaining of the IT systems and platforms Nokia is running on. Develops and maintains applications to meet the business needs of the organisation. Covers the performance and monitoring of day-to-day database and system activities as well as the provision of remote support to users regarding hardware and software problems.

IT Architecture (IAM) comprises the definition and management of enterprise architecture (including solution, security, process & information, and integration architecture), which aligns with and supports Nokia's business strategy, mode of operation, and business needs.

How You Will Contribute And What You Will Learn

Architect agentic AI systems — single-agent, multi-agent, and orchestrator/worker patterns — for autonomous and semi-autonomous DO and business workflows.

Design solutions using modern agentic frameworks such as Microsoft Agent Framework (MAF), Semantic Kernel, LangChain / LangGraph, AutoGen, Google Agent SDK

Define agent capabilities, tool/function calling contracts, memory strategies (short-term, long-term, episodic), planning and reflection loops, and inter-agent communication protocols (e.g., MCP, A2A).

Own the end-to-end architecture: data ingestion, retrieval (RAG, GraphRAG, hybrid search), model selection, agent orchestration, integration, and serving.

Produce high-level and detailed design artifacts: solution blueprints, agent topology diagrams, sequence flows, API/tool contracts, and non-functional requirement specs.

Drive build-vs-buy decisions and evaluate vendor/hyperscaler offerings against Nokia standards.

Ensure solutions align with Nokia's enterprise architecture, security, identity, data-protection, and integration standards.

Define reusable architectural patterns and reference designs for agent-based IT/DO solutions.

Provide hands-on technical leadership across multiple parallel initiatives; mentor GenAI engineers, and platform developers on agentic patterns and frameworks.

Lead technical design reviews, architecture boards, and proof-of-concept evaluations.

Set coding, model, prompt, agent, and AgentOps standards; champion responsible AI, evaluation rigor, and observability.

Define and evolve the GenAI and agentic platform stack — model gateways, vector stores, orchestration frameworks, tool registries, agent runtime, evaluation pipelines, tracing, and cost controls.

Establish AgentOps / LLMOps practices: CI/CD for agents, prompts, and tools; A/B testing; trajectory and trace analysis; drift and quality monitoring; retraining; and decommissioning.

Implement agent observability with frameworks such as LangSmith, OpenTelemetry GenAI, Azure AI Foundry tracing, or equivalents.

Optimize for cost, latency, throughput, and reliability at enterprise scale, including token budgeting and agent step caps.

Translate business problems into agentic solutions; explain trade-offs (autonomy vs control, cost vs capability) in business language to sponsors and leadership.

Engage with hyperscaler partners (Azure, AWS, GCP), SIs, and product vendors on roadmap, capabilities, and joint engineering.

Represent the team in enterprise architecture forums, security reviews, and AI governance councils.

Embed responsible AI by design: bias and fairness checks, content safety, privacy, explainability, and compliance with Nokia's AI policy and regional regulations.

Define agent safety controls — tool-use guardrails, sandboxing, permission scopes, human-in-the-loop approvals, action audit trails, and prompt-injection defenses.

Define evaluation criteria, acceptance gates, and post-deployment monitoring for every AI/agent solution shipped.

Key Skills And Experience

Must-Have:

10+ years in software/solution architecture, with 5+ years architecting AI/ML or data-intensive systems and recent hands-on experience designing GenAI and agentic AI solutions.

Deep, hands-on expertise with at least two agentic frameworks — Microsoft Agent Framework (MAF), Semantic Kernel, LangChain / LangGraph, AutoGen, CrewAI, or LlamaIndex agents.

Strong understanding of agent design patterns: ReAct, Plan-and-Execute, reflection, tree-of-thought, multi-agent collaboration, orchestrator/worker, and supervisor-critic loops.

Proficiency in tool/function calling, structured outputs, MCP, and agent-to-agent communication patterns.

Deep expertise in Python and /or C#.

Strong experience with cloud AI platforms — Azure AI / Azure OpenAI / Azure AI Foundry preferred, plus GCP Vertex AI.

Proven track record designing production-grade RAG, agentic, and fine-tuning solutions, including vector databases (Azure AI Search, Pinecone, Weaviate, pgvector).

Solid foundation in distributed systems, microservices, APIs, event-driven architecture, containers (Docker/Kubernetes), and CI/CD.

Strong grasp of MLOps / LLMOps / AgentOps, evaluation frameworks, and AI observability.

Nice-To-Have:

Excellent communication skills — able to influence both engineering teams and executive stakeholders.ot/cause analysis in more complex problems. Can develop and implement recommendations. Managerial/Supervisory: Direct supervisory responsibilities for people.

Typically first level (and lowest level) of solid line management. Carries out variety of complex activities according to plan within broader area of responsibility, analyses problems. Decision-making typically according to established solutions.