I ship production agentic AI systems — on-prem stacks, multi-agent orchestration, RAG pipelines, MCP servers, with real latency and observability budgets.
- 🔭 Currently Senior Software Engineer at SELISE Group, architecting on-premise Enterprise GenAI ecosystems.
- 🧠 I specialize in LangGraph, RAG (Qdrant + LangChain), multi-agent systems, and MCP for tool-using agents.
- 🎯 Three years shipping LLM systems where hosted APIs weren't an option. ~15% inference-latency wins. Sub-3s streaming for 30k+ users. Multi-million-msg/day IoT middleware.
- 📚 Two published papers in computer vision (saliency-based fish classification at 100% accuracy; Parkinson's classification at 96.67%).
- ✍️ Writing the engineering posts I wish existed — debunking pieces, architecture teardowns, field notes from production.
- 🌍 Portfolio: rafsunjanyarman.com
- 💼 Open to senior / staff Agentic AI roles · Remote-friendly · EU / US / CH time zones.
- Agentic AI: LangGraph, LangChain, CrewAI, MCP (Model Context Protocol), Multi-Agent Systems
- RAG & Vector: Qdrant, hybrid search, chunking strategies, retrieval eval
- Real-time & Async: FastAPI, WebSockets, SignalR, Redis-based distributed cancellation
- Voice AI: LiveKit, Whisper, STT/TTS pipelines
- IoT: Spring Boot, ThingsBoard, MQTT, RabbitMQ, TimescaleDB
- Mobile / Web: Flutter (BLoC, Riverpod), React (Zustand)
- We Migrated from SSE to WebSockets. Cancellation Wasn't the Real Reason. in Level Up Coding
- Is CrewAI Really Async? Debunking the Asynchronous Myth & When Streaming Makes Sense — gitconnected
- Fortress of Flutter: Mastering Testing with Mocktail, Bloc Test, and Firebase — gitconnected
- Automate Flutter Unit Testing & Coverage with SonarQube — Medium
"Three years shipping production agentic AI. Now writing the war stories the tutorials don't cover."