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

