Senior SAP Developer & AI/ML Engineer with 8+ years building scalable enterprise applications. Currently designing production-grade Agentic AI systems integrated with SAP Joule and A2A protocol on BTP/Kyma — delivering token-optimized, scorecard-compliant intelligent automation for Asset Management.
Frameworks, tools, and servers I've built and maintain.
Production-grade AI agent framework with ReAct, PEOS, HITL patterns, tool registry, memory management, and streaming.
Plan-Execute-Observe-Synthesize as a composable LangGraph state machine. Backbone of SAP maintenance agents.
Multi-agent orchestration — supervisor, router, and swarm topologies with shared message bus and delegation.
MCP server exposing any OData V2/V4 service as AI tools. Auto-discovers entities, handles SAP auth.
Kubernetes cluster management via AI assistants. Pods, deployments, logs — via natural language.
MCP server for Confluence wiki search and content retrieval. Query pages, spaces, and attachments.
MCP server for Grafana dashboard queries and alerting insights for observability.
Real-time token cost monitoring and budget enforcement middleware for LLM applications.
Agent-to-Agent protocol toolkit for multi-agent communication, discovery, and task delegation.
Adaptive cruise control via deep reinforcement learning — PPO, DQN, and custom reward shaping.
Turn any OData service into an AI-queryable tool — zero boilerplate, automatic schema introspection.
Swarm intelligence simulation — PSO, ACO, Firefly, and hybrid optimization with visualization.
Prompt engineering toolkit with version control, A/B testing, automatic optimization, and analytics.
Safety middleware for AI agents — prompt injection detection, output validation, and audit trails.
Deep dives on agent architecture, MCP servers, and enterprise AI.