Senior Agentic Engineer

Project Description: Modernization as a Service (MaaS) is a horizontal practice, supporting all business verticals in the company. We are looking for a Senior Agentic Engineer who will be able to work with different modernization projects. Responsibilities: Launch modernization toolset Perform initial result validation Customize and adjust prompts Troubleshoot, debug, and resolve prompt issues Collaborate with modernization team members to align with project goals and deliver high-quality solutions Mandatory Skills Description: Daily work with coding agents (Claude Code, Cursor, Codex or similar) as your main way of writing code, not an experiment. Understanding of how LLM agents work: context, tools, planning, subagents, typical failure modes. Ability to review agent output quickly and critically, catching errors and hallucinations. 5+ years of professional software development as a full stack developer, with systems taken to production and supported there. Relational databases and SQL; understanding of graph data models (Neo4j) and vector databases. Docker, Kubernetes, Helm: you deploy and debug them yourself. Testing as a habit: unit and integration tests, Testcontainers or similar. Understanding of TDD aspects. Owning an area on your own, from an unclear problem statement to production, without someone else breaking it into tasks. Willingness to make decisions with incomplete information and to own the outcome. Ability to get up to speed quickly in unfamiliar code, technology or business domains. Ability to explain a decision and its limits in writing and in person. Nice-to-Have Skills Description: Experience with LLM provider APIs (Anthropic, OpenAI, Bedrock) or a gateway such as LiteLLM. R&D experience: hypotheses, prototypes, experiments and data-driven decisions on problems with no known answer. Mainframe: COBOL, CICS, JCL, DB2, IMS, PL/I, HLASM, CA Gen, z/TPF; 3270 and z/OSMF. Experience building evals for non-deterministic systems: regression suites, metrics, comparison of models and prompts. Knowledge of threats to agents and the code they write: prompt injection, secret leakage, excessive permissions, poisoned dependencies and MCP servers (OWASP Top 10 for Agentic Applications). Secure work with coding agents. Languages: English: B2 Upper Intermediate … mais