AI & Cloud Native Engineer
Building scalable multi-agent platforms, event-driven architectures, and production infrastructure for LLM systems on Kubernetes.
name: Kaushik Maram
role: Agentic AI Engineer
location: Hyderabad, India
focus:
- Multi-Agent Orchestration & Governance
- Event-Driven AI Platforms on Kubernetes
- RAG Systems & LLM Infrastructure
- Cloud Native Platform Engineering
currently_building: orchestration platform for 25+ agents.
writings: "Our GitHub PR Agent Crashed" — a paper on durable state & async scaling for AI agents on K8s|
AI & Orchestration
|
Cloud & DevOps
|
Data & Backend
|
Monitoring & Observability: OpenTelemetry, Kubeflow
|
📄 Our GitHub PR Agent Crashed Paper on engineering crash-resilient AI agent platforms A GitHub PR Review Agent that failed at 10 users, re-architected into an event-driven platform with Redis queuing, KEDA autoscaling, CloudNativePG checkpointing, and MCP governance on Kubernetes. Achieved 100% crash recovery and scale-to-zero efficiency. |
🌾 Agri-Chatbot: Agentic RAG Multi-agent RAG system for agricultural assistance Agentic RAG backend with LangGraph routing, hybrid retrieval (semantic + BM25 + cross-encoder re-ranking), Pinecone vector DB, and MongoDB conversation memory. 30% accuracy improvement over baseline. |
|
🤖 AI Agent RAG Enterprise-ready AI agent with tool calling LangGraph agent with autonomous tool selection, FAISS semantic retrieval, session management, and production REST API deployed on Render. |
👁️ Diabetic Retinopathy Detection Deep learning for medical diagnosis CNN model for multi-class classification across 5 clinical stages with optimized preprocessing (CLAHE, Gaussian Blur), improving detection by 15%. |
- Core Team, CodeDay Hyderabad — Co-organized 24-hour hackathon for 100+ participants, developed GenAI workshop content
- Hackerabadi, Hackerabad — Organized many Tech sessions reaching 500+ students
- Conferences: KubeDay India 2023, PyCon India 2022, GitHub Constellation 2024, Cloud X AI 2026, CNCF Hyderabad Meetups
"AI magic only works when the infrastructure underneath it is boring, predictable, and designed to fail."
From our paper — and the principle I build by.