Headline

Building multi-agent LLM systems that investigate & resolve Kubernetes incidents | R&D Intern @ Fortinet | LangGraph, MCP, RAG, Kubernetes, Python, FastAPI

About

Five AI agents investigate every Kubernetes incident on our platform before a human opens a dashboard. I'm an AI/ML Engineer-in-training building production-grade multi-agent LLM systems — currently an R&D Intern at Fortinet, where I architect FortiWatch, an automated root-cause analysis platform for Kubernetes. It orchestrates five domain agents (Kubernetes, Prometheus, Elasticsearch, Kafka, FortiMonitor) via LangGraph, runs entirely on self-hosted Ollama models with zero data egress, and grounds every verdict against a three-layer ChromaDB RAG knowledge base — so unsupported claims get caught before they reach an engineer. Before Fortinet: I built NLP and semantic-search pipelines that cut manual QA effort 70% and improved data integrity 40% at InternPe; a document analytics platform that reduced redundant entries 85% while processing 200+ documents/hour; and an LSTM sentiment classifier that hit 91% accuracy across 10,000+ samples. I like systems that are honest about what they don't know — that's most of what claim-grounding is really about. Specialties: LangGraph, Model Context Protocol (MCP), multi-agent orchestration, RAG, LLM evaluation, Kubernetes, observability (OpenTelemetry, Arize Phoenix), Python, FastAPI, vector databases, semantic search. Open to LLM/Agent Systems Engineer and AI Infrastructure roles — happy to talk shop if you're building something similar.

Location

Greater Bengaluru Area

AI stack

claude code

claude

hermes agent

hugging face

langgraph

langsmith

Work experience

Research And Development Intern at Fortinet, Mar 2026 - Present · 7 mos, banglore, Built a production-grade LangGraph multi-agent system to automate Kubernetes incident investigation using iterative reasoning and tool orchestration. Implemented MCP integrations for Kubernetes, Prometheus, Elasticsearch, and system telemetry, enabling autonomous evidence gathering across distributed systems. Designed a causal evidence graph to transform logs, metrics, events, and configurations into structured evidence for deterministic root-cause validation using Ollama and ChromaDB-backed RAG.

Data Science Intern at AI4SEE PVT LTD, Jan 2026 - Mar 2026 · 3 mos, Developed and evaluated machine learning models for an AI-driven product, covering data preprocessing, feature engineering, model training, and performance optimization. Contributed to industrialization of the product, taking models from prototype to deployment-ready.

Intern at InternPe, Jul 2025 - Aug 2025 · 2 mos, India

Education

Bachelor of Engineering - BE, Computer and Information Sciences, General, MVJ College of Engineering, Bangalore, India