For my BNY experiential learning case study, I served as Prompt Engineer for a technical project focused on designing and building an autonomous AI agent capable of detecting, analyzing, and summarizing anomalies in a simulated banking application. My system integrates directly with Prometheus, which continuously collects application metrics. When a safeguard rule is triggered, the alert flows through Alertmanager and into my custom-built AI agent. The agent appends the alert details to a structured system prompt and sends it to my LLM engine (Gemini Flash), which is role-prompted to behave as an expert error analyst. Gemini returns a structured JSON response containing a detailed thought process, error summary, and proposed solutions including actionable code-level fixes. I parse this output into clean, readable Markdown reports, automatically saved with timestamps for traceability and review.
This resulted in an end-to-end autonomous pipeline that: