
Software Engineer
Asama.ai · Jul 2025 – Present
- Developed multi-agent workflows using CrewAl, implementing agent roles, custom tools, and orchestration logic to support infrastructure monitoring and troubleshooting across 500+ servers.
- Engineered and iteratively refined 10+ LLM prompts and intent-based routing logic across 5+ infrastructure data sources, improving query routing and agent response relevance by 30%.
- Integrated LangChain for LLM orchestration and Langfuse for observability, enabling end-to-end prompt tracing and performance monitoring across 1,000+ agent executions, reducing LLM workflow debugging time by 40%.
- Implemented real-time LLM response streaming using Server-Sent Events (SSE) through backend event pipelines and frontend rendering, reducing perceived response latency by 50% through incremental response delivery.
- Developed 10+ storage metric collectors for an infrastructure exporter, enabling Prometheus/VictoriaMetrics ingestion and expanding storage health telemetry coverage across 500+ infrastructure nodes.
- Designed and implemented a Neo4j infrastructure knowledge graph using Cypher queries, modeling 500+ infrastructure nodes and 20+ relationship types to power product dashboards and provide graph-based retrieval for Al-powered natural language queries.
- Improved DBSCAN/HDBSCAN-based infrastructure anomaly detection by refining hardware-aware server clustering, reducing false-positive anomalies by 50% and improving detection reliability across 500+ monitored servers.
- Enhanced rule-based infrastructure fault detection by refining 25+ health metric evaluation rules across 10+ server components, reducing false-positive fault detections by 30%.
- Built a threshold-based infrastructure alerting system using vmalert and Alertmanager, supporting 100+ alert rules across 500+ monitored servers, and developed a notification service routing tenant-configured alerts across 4 channels: Slack, Email, Mattermost, and Google Chat.

