# Lim Chee Kin > AI Solutions Architect at Daxonet AI in Kuala Lumpur, Malaysia. Enterprise AI agents, document extraction, RAG and voice AI, built in Python and TypeScript. ## About Lim Chee Kin is an AI Solutions Architect at Daxonet AI, based in Kuala Lumpur, Malaysia. He designs and ships enterprise AI: conversational agents, document extraction, human approvals, and integration with accounting and business platforms. His work spans retrieval-augmented generation (RAG), real-time voice systems, cloud operations and distributed teams. Before AI, he led enterprise Java teams and delivered banking and healthcare systems. ## Core Skills - Languages and backend: Python, TypeScript, JavaScript, Java, SQL; FastAPI, Pydantic, Node.js, Bun, Spring Boot; REST APIs, webhooks and Server-Sent Events - AI engineering: Claude Agent SDK, MCP, Google Gemini, LiteLLM, OpenRouter, ragbits, Docling; multimodal document extraction, agent tool permissions, hybrid retrieval, Ragas and LLM evaluation - Data and storage: PostgreSQL/pgvector, SQLite, SurrealDB, MongoDB Atlas, MinIO - Frontend: React, Next.js, TypeScript, Vite, Tailwind CSS, React Native and Flutter - Voice and enterprise integration: Pipecat, Gemini Live, Whisper, Kokoro; Teams Bot Framework, WhatsApp, Telegram, Google Workspace, AutoCount and Dynamod - Security: OAuth 2.0 with PKCE, JWT validation, role-restricted agent tools, AES-256-GCM token encryption - Infrastructure and quality: Docker/Compose, Kubernetes, Linux, AWS, GCP/GKE, GitHub Actions, pytest and Vitest, OpenTelemetry, Prometheus and Grafana ## Projects At Daxonet AI: - [Claim Pilot](https://kin.rag.wtf/projects/claim-pilot/): Expense claims from Teams and WhatsApp, read by AI, approved by humans, and posted as GL journals. - [Trade Pilot](https://kin.rag.wtf/projects/trade-pilot/): Purchase orders arrive on WhatsApp as photos, PDFs or Excel, and leave as quotations and sales orders. - [WareBot](https://kin.rag.wtf/projects/warebot/): Ask the warehouse in plain language. Every write waits for an explicit confirmation. - [Agentic AI Layer](https://kin.rag.wtf/projects/agentic-ai-layer/): Multi-channel agents that reach accounting and business systems through role-restricted MCP tools. - [Smart HVAC Catalogue System](https://kin.rag.wtf/projects/smart-hvac-catalogue/): Smart HVAC Catalogue System: a RAG backend that matches air-conditioners to room type, size and usage with hybrid retrieval and product filters. Open source: - [Project Concord](https://kin.rag.wtf/projects/project-concord/): An MCP server that lets modern AI agents query legacy database schemas, with permission controls. Source: https://github.com/limcheekin/project-concord - [Talk To AI](https://kin.rag.wtf/projects/talk-to-ai/): Real-time voice AI targeting sub-300 ms responses, with local or cloud speech and models. Source: https://github.com/limcheekin/talk-to-ai - [RAG.WTF](https://kin.rag.wtf/projects/rag-wtf/): Private, personalised knowledge discovery: your documents, your models, your machine. Source: https://github.com/rag-wtf - [Fluwix](https://kin.rag.wtf/projects/fluwix/): Fluwix is an open-source showcase of Flutter applications by Lim Chee Kin, exploring what the Flutter framework can do on mobile. Source: https://github.com/limcheekin/fluwix - [open-text-embeddings Python Library](https://kin.rag.wtf/projects/open-text-embeddings/): open-text-embeddings is an open-source Python library that serves BGE, E5 and other sentence-transformer models behind an OpenAI-compatible API. Source: https://github.com/limcheekin/open-text-embeddings - [Self-Hosted AI Infrastructure](https://kin.rag.wtf/projects/self-hosted-ai-infra/): Self-hosted AI infrastructure by Lim Chee Kin: local LLMs, embeddings, reranking and speech services on an OrangePi 5 Max with Docker and Coolify. - [Talking Book](https://kin.rag.wtf/projects/talking-book/): Talking Book is a YouTube channel by Lim Chee Kin that turns non-fiction, business and self-improvement books into conversations, summaries and songs. Source: https://github.com/limcheekin/talking-book - [TelegramGPT](https://kin.rag.wtf/projects/telegram-gpt/): TelegramGPT is an open-source Telegram bot on Google Gemini, with PostgreSQL conversation history and optional self-hosted speech services. Source: https://github.com/limcheekin/TelegramGPT ## Past Contributions - 10+ Grails Framework plugins for the developer community ## Work Experience - 2026-Present: AI Solutions Architect (Daxonet AI) - 2021-2025: Principal AI Architect (Independent Consultant) - 2013-2020: Senior Java Tech Lead (DXC Technology) - 2012-2013: Senior Java Tech Lead (Zurich Technology Services Malaysia) - 2010-2012: Independent Software Professional - 2008-2010: Senior Solution Consultant (Penril Datability) - 2005-2007: Technical Lead (CMG Online) - 2004-2005: Founder and Senior Consultant (Media Trend) - 2002-2004: Software Engineer (Siemens Multimedia) - 2001-2002: System Engineer (Online One Corporation) - 1999-2000: System Engineer (Tee Yam Holding) ## Articles - [Zero-Trust Security for Autonomous AI Agents](https://limcheekin.medium.com/zero-trust-security-for-autonomous-ai-agents-691b5cbcec21) - [Solving the LLM Load Testing Paradox: A Reference Architecture on GKE Autopilot](https://limcheekin.medium.com/solving-the-llm-load-testing-paradox-a-reference-architecture-on-gke-autopilot-c8c2271f5cfb) - [Reproducing Karpathy's NanoChat on a Single GPU — Step by Step with AI Tools](https://limcheekin.medium.com/reproducing-karpathys-nanochat-on-a-single-gpu-step-by-step-with-ai-tools-e9420aaee912) - [AI is Ready. Your Legacy Database Isn't. Let's Fix That.](https://limcheekin.medium.com/ai-is-ready-your-legacy-database-isnt-let-s-fix-that-f28013a2f833) - [Building Your First No-Code MCP Server: The Fabric Integration Story](https://limcheekin.medium.com/building-your-first-no-code-mcp-server-the-fabric-integration-story-90da58cdbe1f) - [Introducing Talking Book](https://limcheekin.medium.com/introducing-talking-book-487f6e3bc2c2) - [Beyond the Cloud: How I Built My Own AI Server (and Why)](https://limcheekin.medium.com/beyond-the-cloud-how-i-built-my-own-ai-server-and-why-68b7235117f3) - [Would you pay $1/month to Own Your AI Data?](https://limcheekin.medium.com/would-you-pay-1-month-to-own-your-ai-data-6dbb0db1eeaf) - [Talk To AI with FastRTC: Real-Time Voice Conversations with AI](https://limcheekin.medium.com/talk-to-ai-with-fastrtc-real-time-voice-conversations-with-ai-5cf35f9691d8) - [Unleashing the Power of Open Source Text Embeddings with open-text-embeddings](https://limcheekin.medium.com/unleashing-the-power-of-open-source-text-embeddings-with-open-text-embeddings-e8fc684aa2cd) ## Contact - Email: limcheekin@vobject.com - GitHub: https://github.com/limcheekin - LinkedIn: https://linkedin.com/in/limcheekin - Medium: https://medium.com/@limcheekin ## Technical - Portfolio: https://kin.rag.wtf/ - Full Content (llms-full.txt): https://kin.rag.wtf/llms-full.txt - Sitemap: https://kin.rag.wtf/sitemap-index.xml - Portfolio Data (JSON): https://kin.rag.wtf/api/portfolio.json - OpenAPI Spec: https://kin.rag.wtf/api/openapi.json - Plugin Discovery: https://kin.rag.wtf/ai-plugin.json - A2A Agent Card: https://kin.rag.wtf/.well-known/agent.json (also at /.well-known/agent-card.json) - MCP Server (read-only, Streamable HTTP, POST only, no auth): https://kin.rag.wtf/mcp — tools: get_projects, get_project, get_experience, get_articles, search_portfolio, get_contact_info - MCP Manifest: https://kin.rag.wtf/.well-known/mcp.json - Per-collection JSON: https://kin.rag.wtf/api/projects.json, https://kin.rag.wtf/api/experience.json, https://kin.rag.wtf/api/articles.json (index of all endpoints: https://kin.rag.wtf/api/) - Search index (GET, no MCP needed): https://kin.rag.wtf/api/search.json — every project, article and role with keywords and page URL - RSS (Writing): https://kin.rag.wtf/rss.xml - Security contact: https://kin.rag.wtf/.well-known/security.txt - Agent Policy: https://kin.rag.wtf/.well-known/agent-policy ## Site Map Notes - The navigation label "Work" opens /projects/ and "Writing" opens /insights/. - Project pages: https://kin.rag.wtf/projects// ; article summaries: https://kin.rag.wtf/insights// (slugs are listed in /api/portfolio.json). - Markdown versions: append index.md to a project, article or About URL (e.g. https://kin.rag.wtf/projects/claim-pilot/index.md, https://kin.rag.wtf/about/index.md); each page links its own with . - Roles on About have anchors: https://kin.rag.wtf/about/#role- (e.g. #role-dxc-tech-lead).