garvit = {
"role": "GenAI Builder + Agentic AI Architect",
"city": "Delhi, India 🇮🇳",
"focus": ["LLM Orchestration", "Multi-Agent Systems", "RAG"],
"stack": ["LangGraph", "CrewAI", "LlamaIndex", "FastAPI", "React"],
"obsession": "AI that does things, not just says things",
"online": "24/7 when the idea hits 🌙",
"endgame": "build an AI that builds AI",
}
🧠 Orchestration
🤖 Models & APIs
📦 Vector & Memory
⚛️ Frontend
🔧 Backend & Infra
🎨 Tools & Design
| Topic | Why it matters to me | |
|---|---|---|
| 🕸️ | Multi-Agent Systems | Agents that spawn, coordinate & self-improve |
| 🔁 | LangGraph Workflows | Stateful, looping AI pipelines with real memory |
| 🛠️ | Tool-Calling LLMs | AI that acts, not just chatters |
| 📚 | RAG at Scale | Knowledge bases that actually answer right |
| 🔌 | MCP Protocol | The next standard for AI tooling |
| ⚡ | Streaming Agentic AI | Live decisions, zero wait |
| 🔬 | AI Evals & Observability | Vibes aren't a metric |
"Most devs use AI as a search engine. I use it as a teammate."
"Your RAG pipeline is only as good as your chunking strategy. Fight me."
"The best UI is the one users forget they're using."
"Agentic AI isn't the future — it's already late."
╔══════════════════════════════════════════════════════════════╗
║ ║
║ unpopular opinion: context windows matter more than ║
║ model size. a well-prompted small model beats a lazy ║
║ GPT-4 prompt every single time. ║
║ ║
║ — Garvit, debugging a LangGraph node at 2am ║
╚══════════════════════════════════════════════════════════════╝
