AI Architect (Polish is Mandatory) (She/He/They)

CAPCO POLAND

Architect AI

Location: Warsaw, Poland

Pref. work model - 3x per week from office

ROLE OVERVIEW

We are looking for AI Architect that leads the design and evolution of AI agent architecture within the Master Agent Knowledge Management ecosystem and domain-specific agents. The role ensures secure, scalable, and well-integrated solutions that enable consistent knowledge flow and high-quality agent performance. By combining technical expertise with architectural and product thinking, the AI Architect drives the stability, compliance, and continuous improvement of the AI agent platform in line with business needs.

Fluency in Polish is mandatory.

RESPONSIBILITIES

• Designing AI agent architecture (Master Agent + domain agents) in Copilot Studio, Azure, and multicloud integrations (Azure/GCP).
• Modeling agent workflows: orchestration, task delegation, context handling, agent registry, and profiling instructions.
• Designing and implementing integrations: REST/APIs, MCP, custom tools, custom connectors, and Graph Connectors.
• Designing and maintaining knowledge layers: Azure AI Search / Knowledge Bases, indexes, metadata, RAG policies, and cataloging standards (SharePoint Online).
• Designing and overseeing conversational UX: prompt engineering, intent design, dialog flows, and context interpretation.
• Defining agent quality metrics: factuality, relevance, compliance, hallucination rate, coverage, latency, and response cost.
• Designing and implementing LLMOps: automated evaluations, test pipelines, monitoring, observability, and alerting.
• Managing security: PII compliance, anonymization/masking, access control (Entra ID), and document/segment-level authorization.
• Optimizing performance: latency, token usage, caching, vector policies, and inference cost per query.
• Coordinating integrations with other agents and platforms (Power Platform, Foundry AI, GCP BigQuery/Looker, Vertex AI, Document Intelligence).
• Defining architecture standards and integration patterns for the entire Knowledge Management program.
• Collaborating with DevOps, Security, AI/ML Engineers, and domain owners to ensure compliance with implementation standards.

SKILLS & EXPERIENCES TO GET THE JOB DONE

• Very good knowledge of Azure OpenAI / Azure AI Foundry and Azure AI Search / Knowledge Bases.
• Hands-on experience with Copilot Studio (agents, prompt flows, tools, integrations).
• Strong knowledge of Azure, Log Analytics, and observability practices for AI agents.
• Understanding of Entra ID: permissions, OAuth2, authorization flows, and document/index-level access control.
• Experience with GitHub Actions, Azure Container Registry (ACR), and CI/CD pipelines.
• Ability to design event-driven architectures (Service Bus, Event Grid, Durable Functions).
• Experience with REST/API integrations, webhooks, Graph API, and MCP.
• Knowledge of integration patterns in agent-based architecture and multicloud environments (Azure ↔ GCP).
• Practical experience with LLMOps: quality evaluations, regression testing, and guardrails.
• Good understanding of RAG principles, metadata processing, and indexing policies.
• Strong knowledge of data security, including PII masking and regulatory compliance.
• Ability to design AI agents in multi-agent ecosystems (Master Agent + agent-to-agent communication).

Nice to have:

• Experience with Microsoft Fabric / Fabric AI (Lakehouse, OneLake, semantic models).
• Experience with Graph Connectors (M365, SharePoint, Teams).
• Experience with Azure Document Intelligence / OCR.
• Python skills (LLM pipelines, automated testing).
• Docker, AKS, and Azure Container Apps (ACA).
• Durable Functions and Logic Apps.
• Integrations with Microsoft Graph (mailbox, Teams, ChatOps).

IMPORTANT

• Fluent Polish (spoken and written) – mandatory
• Good command of English for documentation and collaboration.
• Availability to work on-site, with partial remote work - 3 days per week from the office.

ONLINE RECRUITMENT PROCESS STEPS

• Screening call with the Recruiter
• Hiring Manager Technical Interview
• Client stage
• Feedback/Offer

We offer a flexible collaboration model based on a B2B contract, with the opportunity to work on diverse projects.