AI-Assisted Application Monitoring Agent
An AI-assisted application monitoring system that allows developers to investigate application activity through natural-language queries.
Technologies & Tools
Problem Statement
Application monitoring systems can generate large amounts of logs and event data. Investigating those events manually requires developers to search through stored records to understand what happened, when it happened, which parts of the application were involved, and what patterns exist in the data.
Role & Contributions
Backend / AI Engineer (Telex Contributor). Built backend services, API endpoints, event storage workflows, AI integration, and connected natural-language queries with stored application telemetry.
System Architecture & Design
Built with C# and ASP.NET Core for structured API development and high-throughput telemetry ingestion, coupled with MongoDB for persistent monitoring event storage. Semantic Kernel coordinates AI workflows and connects natural-language queries with stored telemetry context.
Key Engineering Decisions
- 1.AI as an Investigation Layer: Designed the system so the AI does not monitor the app directly, but accesses stored monitoring events as an intelligent exploration interface.
- 2.C# / ASP.NET Core: Chosen for high-throughput asynchronous execution, low memory overhead, and strong architectural boundaries.
- 3.MongoDB: Selected for flexible storage and rapid querying of semi-structured application logs and telemetry events.
- 4.Semantic Kernel: Coordinated AI workflows and linked language model capabilities with backend data retrieval functions.
Challenges Overcome
- Connecting unstructured natural-language developer questions with relevant structured monitoring event records.
- Curating high-signal context from high-volume telemetry to deliver concise, actionable diagnostic summaries.
Lessons Learned
- AI features are most effective when grounded directly in real application data rather than operating in isolation.
- Context quality and targeted retrieval strongly govern the accuracy and usefulness of AI diagnostic responses.
Results & Impact
- Successfully integrated into the Telex collaborative agent platform.
- Enabled developers to query complex server logs, errors, and system telemetry through conversational interactions.