📍 Tokyo, Japan | 🤖 GenAI Engineer | ☸️ CKA | 🏗️ Ex-OpenStack Contributor

GenAI engineer with 10+ years in cloud-native infrastructure, Python, and distributed systems — now focused on building production-grade LLM pipelines, RAG systems, and AI agents that are reliable, observable, and safe to deploy.

Currently SDE at Mercari (GKE platform, Terraform, DORA metrics). Previously Rakuten Mobile (bare-metal K8s for 5G).

Upstream contributor to OpenStack (2017–2020) — identity, authentication, and compute features shipped to production clouds worldwide.

• 🔐 openstack/keystone — Core contributor; token federation, identity mapping, and auth policy
• ☁️ openstack/nova — Compute platform contributions
• 🌍 openstack/openstack — Cross-project integration contributions

🎤 Open Infrastructure Summit presenter · Denver · Shanghai

• 🤖 multi_agent_coding_assistant — Planner → Coder → Reviewer agents for end-to-end Python tasks
• 🔍 agentic_rag_system — Self-reflective LangGraph RAG with query rewriting and grading
• 🕸️ graphrag_knowledge_system — Neo4j knowledge graph + local/global GraphRAG search
• 🏥 offline_medical_agent — Fully offline clinical protocol lookup with agentic RAG

• 🤖 multi_agent_coding_assistant — Multi-agent code generation with planner, coder, and reviewer
• 📊 langchain_data_agent — Natural language SQL over SQLite with read-only enforcement
• 📚 documentation_qna_agent — Chat with any documentation URL
• 🎯 competitive_intelligence_agent — CrewAI-powered sales battlecard generator
• 💰 personal_finance_agent — Bank statement Q&A with tool-calling agents
• 🌐 browser_automation_agent — Autonomous web navigation from plain English

• 🎬 youtube_transcript_rag — Chat with YouTube videos via transcript RAG
• 🧠 reasoning_rag — RAG with visible step-by-step reasoning traces
• 📄 hyde_rag — Hypothetical Document Embeddings for better retrieval
• 🔄 agentic_rag_system — Self-reflective LangGraph RAG pipeline
• 🕸️ graphrag_knowledge_system — Document ingestion → entity extraction → graph + vector search
• 🎨 multimodal_rag — RAG over text, PDFs, images, audio, and video via Gemini File API

• 🧾 receipt_expense_tracker — Offline receipt OCR with Gemma vision and expense ledger
• 🎙️ multilingual_audio_translator — Speech-to-speech translation pipeline
• 📹 video_understanding_agent — YouTube video summarization with Gemini
• 🏥 offline_medical_agent — Offline clinical protocol lookup with Qdrant + local LLM

• Production-grade AI — Building LLM systems with the same rigor as platform engineering: reliability, observability, and guardrails
• RAG at scale — Hybrid retrieval, graph-augmented search, and multimodal indexing over real-world document corpora
• Multi-agent orchestration — LangGraph and CrewAI pipelines for complex, tool-using workflows
• Infrastructure meets AI — Applying a decade of K8s, Terraform, and distributed systems experience to AI deployment

Certifications: Certified Kubernetes Administrator (CKA)

📫 agarwalvishakha18@gmail.com · LinkedIn

Infrastructure engineers know how to make things run at scale. I'm applying that mindset to GenAI — because the hardest part of AI isn't the demo, it's making it reliable in production.