"The universe is 13.8 billion years old. I've been building AI for a fraction of that — but the models are getting smarter."
I'm a Full-Stack AI Engineer specializing in building end-to-end intelligent systems — from raw data pipelines to production-deployed AI products. My work sits at the intersection of LLMs, retrieval systems, and autonomous agents.
class PiyushKumar:
name = "Piyush Kumar"
title = "Full-Stack AI Engineer"
focus = ["RAG Systems", "LLM Finetuning", "Agent Orchestration"]
tools = ["Antigravity 🪐", "Kimi Work 🌙", "VS Code ⚡"]
stack = "Full-Stack AI" # from embeddings to UI
mindset = "ship fast · learn deep · iterate relentlessly"
current_era = 2026
def greet(self):
return "Welcome to my corner of the AI universe 🌌"- Hybrid search · vector stores · reranking pipelines
- Context-aware chunking · multi-modal retrieval
- Production RAG with evals & observability
- Instruction tuning · LoRA / QLoRA · RLHF / DPO
- Domain adaptation · dataset curation & synthetic data
- Model evaluation · MMLU, custom benchmarks
- Multi-agent frameworks · tool-use · planning loops
- Memory systems · long-horizon task execution
- Agentic pipelines with guardrails & fallbacks
| Layer | Tools |
|---|---|
| AI Dev IDE | Antigravity 🪐 · Kimi Work 🌙 |
| Editor | VS Code ⚡ |
| LLM Frameworks | LangChain · LlamaIndex · HuggingFace Transformers |
| Languages | Python 🐍 |
| Infra | Git & GitHub 🌍 |
🔹 Production RAG Pipeline – Multi-stage retrieval with hybrid search, reranking & LLM synthesis
🔹 Finetuned Domain LLM – Custom instruction-tuned model with LoRA for specialized Q&A
🔹 Multi-Agent Workflow – Orchestrated agents for autonomous task decomposition & execution
"The best model is the one that ships." 🚀
Full-Stack AI · RAG · Finetuning · Agent Orchestration