Oppex AI: Automating Production Incident Resolution with AI Agents
How Oppex AI's agents reduce MTTR by gathering context before waking developers
1AI-powered incident resolution
Oppex AI develops autonomous agents that diagnose and recommend fixes for production incidents before developers are alerted. The system aggregates logs, cloud metrics, database health, affected customers, and recent code changes from GitHub. By analyzing historical data, it identifies whether an issue has occurred before and how it was resolved. This contextual synthesis aims to reduce mean time to resolution (MTTR) below 10 minutes, significantly faster than traditional manual debugging processes.
2Chaos engineering approach
The team trains their AI agents using a 'chaos monkey' methodology, where test systems are intentionally broken without informing the agent. This adversarial training strengthens the system's reliability by forcing it to learn from failures. In production environments with 5,000+ machines and 250+ microservices, the AI builds a knowledge graph from logs that aren't in plain English. This approach helps the system understand complex distributed architectures without human intervention.
3Technical implementation stack
90% of Oppex AI's development occurs in JetBrains' IntelliJ IDEA due to their Java-heavy codebase, supplemented by WebStorm for JavaScript, DataGrip for database layers, and PyCharm for Python components. The team integrates JetBrains AI Assistant throughout their workflow. This IDE-centric approach enables rapid code navigation and dependency analysis, critical for maintaining complex AI systems. The founder has used JetBrains tools continuously since 2001, demonstrating their long-term viability for serious engineering work.
4Production context assembly
When an incident occurs, Oppex AI's agents automatically gather diagnostic information from multiple sources before alerting engineers. They analyze cloud metrics to identify resource constraints, examine database load patterns, and determine customer impact through live product monitoring. The system cross-references GitHub change logs to identify potentially problematic recent modifications. All this context is compiled into actionable recommendations delivered directly to developers' IDEs.
5Founder's engineering background
Prasun Kumar brings 23 years of startup experience to Oppex AI, having led engineering at six previous ventures that achieved successful exits. His technical focus has consistently been on the intersection of product development and scalable systems. This seventh startup reflects his accumulated knowledge about the challenges of maintaining production reliability while moving quickly - a tension he's encountered throughout his career in scaling engineering organizations.
6Future development roadmap
The team's immediate goal is achieving consistent sub-10-minute MTTR with human oversight. Next, they aim for 'AI-recommended, human-approved' resolutions requiring just one-click approval. The long-term vision involves fully autonomous systems that detect, diagnose, and repair issues without human intervention. This evolution mirrors the industry's broader shift toward self-healing systems as AI capabilities mature and gain operational trust.
7Technical challenges overcome
Two significant hurdles required innovative solutions: interpreting non-English developer logs and hardening the AI's decision-making. The team developed specialized parsing to transform technical logs into structured knowledge graphs. For reliability, they created simulated production environments where the chaos monkey constantly introduces novel failures. This combination of semantic understanding and stress testing produces agents capable of handling real-world complexity.
8Strategic industry positioning
Oppex AI enters a market where AI-generated code is increasing production incidents while engineering teams manage more code than ever. Their solution addresses the critical gap between rapid development and reliable operations. As Kumar notes, existing categories are ripe for disruption by AI-first approaches, as larger organizations struggle to adapt quickly. The startup's focus on production reliability positions it well in the expanding AIOps sector.