Senior AI Workflow Development Engineer

We are seeking a Senior AI Workflow Development Engineer to build agentic workflows that support exception handling, data classification, decision-making and reporting automation. This role requires strong hands-on application development experience using AI tools and frameworks, with a focus on delivering practical workflow automation in partnership with operations teams.

Please note that working from the customer's office is required 4 days per week.

Responsibilities

Build AI-powered workflows for exception handling, triage, categorization and decision support Develop agents able to process data quality issues, natural-language exceptions and vendor-submitted issue descriptions Create automation for back-reporting processes and operational workflow support Implement natural language matching and classification logic for data and reporting issues Deliver proof-of-concept and production-ready applications using LLM-based approaches Collaborate with Reg Ops and business users to extract workflow steps, decision points and cognitive load from current processes Support eventual rule-text interpretation, sourcing logic, transformation logic and code-generation use cases Help reduce manual effort in onboarding and handling large volumes of reports and exceptions Requirements Hands-on experience building AI applications, not just experimenting with models Proficiency in Lang.

Chain, AWS and LLM-based application development Expertise in implementing RAG solutions Skills in workflow automation and application development Capability to work with operations stakeholders to translate business processes into technical workflows Local presence in Warsaw and ability to work under a local-to-local contract Nice to have Background in exception management, classification or decision automation workflows Familiarity with regulated operations or reporting environments Knowledge of data quality issue handling and vendor issue matching Capability to support agentic roadmap work from initial classification through action-taking workflows