Skip to main content
L

About Lim Chee Kin

I'm Lim Chee Kin, an AI Solutions Architect at Daxonet AI in Kuala Lumpur. I design and ship enterprise AI: conversational agents, document extraction, human approvals, and integration with accounting and business platforms.

My work spans retrieval-augmented generation (RAG), real-time voice systems, cloud operations and distributed teams, built in Python and TypeScript. Before AI, I led enterprise Java teams across China, Europe and Malaysia, and delivered banking and healthcare systems.

I care about AI that does real work safely: role-restricted agent tools, explicit confirmation before writes, and private, self-hosted models where data sovereignty matters.

Technologies I work with

  • Agentic AI & MCP
  • RAG & LLM Evaluation
  • Voice AI (Pipecat)
  • Python & FastAPI
  • TypeScript & React
  • Claude Agent SDK & Gemini
  • Docker & Kubernetes
  • GCP & AWS
Professional headshot of Lim Chee Kin

Work Experience

  1. 2026-01 – Present

    AI Solutions Architect @ Daxonet AI

    • Claim Pilot: Built expense claim intake through Teams and WhatsApp, document capture in a web portal, and claimant, verifier and approver reviews. Implemented Gemini receipt extraction, a FastAPI backend and React portal, with accounting/ERP integration for invoices and general ledger journals.
    • Trade Pilot: Developed WhatsApp intake for purchase orders, delivery orders and item lists from messages, images, PDFs and Excel files. Combined document extraction with a Next.js review portal and AutoCount/Dynamod integration for quotations, sales orders and goods received notes, supporting admin review or automatic processing.
    • WareBot: Delivered a conversational AI assistant for Warehouse Management System with natural-language stock queries, warehouse operations, SSE streaming and explicit confirmation before writes. Integrated locally hosted Qwen3.5 inference and delivered three Maestro UI test flows and a 21-assertion Newman API suite.
    • Smart HVAC Catalogue System: Built a RAG backend matching air-conditioners to room type, size and usage. Combined PDF ingestion, dense and sparse retrieval, and product filters; implemented asynchronous ingestion, LLM evaluation and observability, with tests and evaluations in GitHub Actions.
    • Agentic AI Layer: Architected multi-channel agents connecting accounting and business systems through APIs and role-restricted MCP tools. Implemented Claude Agent SDK workflows, Gemini-backed memory, Pipecat voice, and Telegram/WhatsApp access; integrated Google Workspace and Microsoft Graph using OAuth with PKCE and encrypted token storage.
    • Agent First Web Audit: Developed a pipeline of four agent skills to audit websites for AI readiness, apply fixes and review changes.
    • AI Enablement: Conducted company-wide training for functional consultants, engineers and support teams, plus monthly AI Ask Me Anything sessions for practical sharing and Q&A.
    • Python
    • TypeScript
    • FastAPI
    • React
    • Next.js
    • Claude Agent SDK
    • MCP
    • Google Gemini
    • RAG
    • Pipecat
    • OAuth 2.0 with PKCE
    • GitHub Actions
    • Solution Architecture

  2. 2021-01 – 2025-12

    Principal AI Architect @ Independent Consultant

    • Launched Project Concord, a TypeScript-based, Dockerized Model Context Protocol (MCP) server that bridges modern AI agents with legacy databases by translating ambiguous schemas into clear, AI-friendly interfaces.
    • Championed and developed open-source AI solutions focusing on data ownership, privacy, and efficient local AI processing; architecting and building RAG.WTF as a platform for secure, personalized knowledge discovery and management.
    • Engineered 'open-text-embeddings', a Python library providing an OpenAI-compatible API for diverse open-source sentence transformer models, enabling significant cost reduction and eliminating vendor lock-in.
    • Developed 'Talk To AI' application, integrating HuggingFace's FastRTC for low-latency real-time voice AI interactions; engineered a flexible backend supporting local and cloud-based STT, LLM, and TTS APIs.
    • Pioneered and operationalized a self-hosted AI infrastructure (Coolify, Docker), running multiple local LLMs (Llama 3.2, DeepSeek-R1, etc.) and STT/TTS for data privacy and operational control.
    • Advocated for and implemented robust data sovereignty by integrating embedded database solutions like SurrealDB.wasm with RAG.WTF.
    • Strategically utilized TensorDock and Modal for scalable cloud GPU servers and serverless execution AI/ML workloads, complementing local AI capabilities.
    • Previously initiated Talking Book YouTube channel (AI-powered summaries and Telegram bot) and Fluwix (Flutter showcases).
    • Deployed a healthcare agent on GKE using Vertex AI and MongoDB Atlas with Private Service Connect; configured Cloud Logging, latency/error alerts and horizontal pod autoscaling.
    • Integrated Ragas to evaluate retrieval accuracy and generation quality, providing repeatable checks for RAG outputs.
    • Open-Source AI
    • RAG
    • MCP
    • Python
    • Text Embeddings
    • FastRTC
    • Docker
    • Local LLMs
    • SurrealDB.wasm
    • Modal.com
    • FastAPI
    • Flutter
    • Voice AI
    • Solution Architecture
    • GKE
    • Vertex AI
    • MongoDB Atlas
    • Ragas

  3. 2013-01 – 2020-12

    Senior Java Tech Lead @ DXC Technology

    • Led a 6-developer international team (China, Europe, Malaysia) to deliver the Analytic Console for the Agile Process Automation (APA) platform (VueJS, Spring Boot, GoLang, MongoDB, AWS), contributing millions to company revenue.
    • Headed a 5-developer international team (US, Malaysia) to create the EPIC Configurator for internal Enterprise Invoice Processing (JQuery UI, Grails Framework, Oracle DB), significantly streamlining configuration processes.
    • Implemented a Divestiture and Acquisition (DnA) dashboard (AngularJS, Jersey, Spring, MongoDB, AWS) for critical real-time data insights supporting strategic executive decision-making.
    • Prototyped an Intelligent Voice Agent (Conversational AI) leveraging Amazon Connect, AWS Lambda, Amazon Lex, Dialogflow, and Micronaut Framework.
    • Established code quality gates and design reviews, and mentored three junior developers.
    • Java
    • Spring Boot
    • VueJS
    • GoLang
    • MongoDB
    • AWS
    • JQuery UI
    • Grails Framework
    • Oracle DB
    • AngularJS
    • Micronaut
    • Conversational AI
    • Team Leadership

  4. 2012-01 – 2013-12

    Senior Java Tech Lead @ Zurich Technology Services Malaysia

    • Led a 5-member production support and development team for motor and property claims systems (IBM WebSphere, FileNet, DB2, Spring, GWT, Hibernate/JPA).
    • Resolved over 60 critical production issues within weeks, dramatically improving system stability and user satisfaction.
    • Strengthened testing and operational safeguards across the claims systems.
    • Java
    • IBM WebSphere
    • FileNet
    • DB2
    • Spring
    • GWT
    • Hibernate/JPA
    • Production Support
    • Team Leadership

  5. 2010-01 – 2012-12

    Independent Software Professional @ Self-Employed

    • Developed and maintained impactful open-source Grails Framework projects, including BPM workflow, form builder, and validation components.
    • Grails Activiti Plugin: BPM workflow system; grew the forum to 100+ members, addressed 100+ queries.
    • Grails Form Builder Plugin: Empowered non-programmers to create online forms.
    • JQuery Validation UI Plugin: Enhanced developer productivity with client-side validation.
    • Explore all 10+ projects at limcheekin.blogspot.com/p/my-grails-plugins.html.
    • Grails Framework
    • Activiti BPM
    • JQuery
    • Open Source Development
    • Plugin Development
    • Groovy
    • Java

  6. 2008-04 – 2010-07

    Senior Solution Consultant @ Penril Datability (M) Sdn. Bhd.

    • Led the successful tender for a RM6 million Internet Banking System (IBS) for Kuwait Finance House (KFH).
    • Designed and proposed comprehensive IBS solutions, covering 24x7 web layers, external integrations (Payment Gateway, SMS, Email), application platform, 2-Factor Authentication (2FA), Enterprise Service Bus (ESB) for backend integration, and High-Availability infrastructure.
    • Led a 3-member team to develop KFH's information website using BroadVision 8.1 Business Agility Suites and Oracle 10g, including JSR168 compliant portlets like Prayer Times and Currency Converter.
    • Secured a RM2 million eCustody System tender from Malayan Banking Berhad.
    • Provided pre-sales support, product demonstrations, solution consultation, and Proof-Of-Concepts (POCs) to major financial and corporate clients (e.g., Maybank, BSN, BNM, Agrobank, ING, Bank Islam).
    • Solution Consulting
    • Tender Management
    • Banking Systems
    • Solution Architecture
    • BroadVision
    • Oracle 10g
    • JSR168 Portlets
    • BPMN
    • Requirements Analysis
    • Pre-Sales
    • Team Leadership

  7. 2005-04 – 2007-12

    Technical Lead @ CMG Online Sdn. Bhd.

    • Led an 8-member international team (China, India, Malaysia) to deliver a RM2 million Integrated Hospital Inventory System (IHIS4) for government hospitals, using AIX UNIX, JBoss 4.0, Oracle 10g, J2EE, Struts, and JasperReports.
    • Drove Quality Management Systems (QMS): introduced a Standard Development Environment (Eclipse, Subversion, JUnit, Ant), improving team productivity by 20%.
    • Actively participated in the CMMI Committee to define and implement CMMI Level 3 compliant organizational standards and processes.
    • Architected and developed a common technical foundation for the Integrated Hospital Information Systems (IHIS5) on AppFuse (J2EE, Spring Framework, Hibernate, JUnit, JMock), including reusable logging and auditing modules with Spring AOP.
    • Technical Leadership
    • J2EE
    • JBoss
    • Oracle 10g
    • Struts
    • JasperReports
    • Spring Framework
    • Hibernate
    • Spring AOP
    • AppFuse
    • CMMI Level 3
    • QMS
    • SDLC
    • Team Management

  8. 2004-05 – 2005-03

    Founder and Senior Consultant @ Media Trend

    • As Founder, led a startup providing web design, website development, web application development, and printing services.
    • Designed and developed websites for clients including Bright Pancar Enterprise and Marklon Industries Sdn. Bhd.
    • Provided printing services (name cards, brochures, cash sales receipts) for various companies.
    • Entrepreneurship
    • Web Design
    • Web Development
    • Client Management
    • Software Consulting

  9. 2002-10 – 2004-04

    Software Engineer @ Siemens Multimedia Sdn. Bhd.

    • Developed the core Voice4Info Portal (News Delivery over Phone), including Call Control Components using Java IDL, CORBA, Parlay API, and VoiceXML, collaborating with German colleagues on the Siemens Next Generation Network (NGN) platform.
    • Engineered iTalk, a Unified Phone Conferencing system with instant messaging-like features (presence, contact list, buddy list management).
    • Supervised and coordinated 4 groups of Multimedia University (MMU) final year students on IP telephony projects, ensuring functional prototypes for the Siemens NGN platform.
    • Java
    • J2EE
    • CORBA
    • Parlay API
    • VoiceXML
    • Telephony Systems
    • Siemens NGN
    • Software Development
    • Mentorship

  10. 2001-08 – 2002-09

    System Engineer (Web Application Development) @ Online One Corporation Berhad

    • Successfully deployed E-Treasury Management Systems to Hitachi Asia Ltd (Singapore) within 2 months, contributing RM1 million in revenue.
    • Developed Fund Transfer, Bank Reconciliation, and General Ledger modules for the E-Treasury system using Cold Fusion, HTML, JavaScript, JavaBean, JDBC, and Oracle 8i.
    • Engineered Business-to-Business (B2B) integration of National Panasonic of Malaysia's purchasing system using Active Server Page (ASP), SOAP, XML, IIS Web Server, and Oracle 8i.
    • Web Application Development
    • Cold Fusion
    • Oracle 8i
    • ASP
    • SOAP
    • XML
    • HTML
    • JavaScript
    • JDBC
    • System Integration

  11. 1999-06 – 2000-05

    System Engineer @ Tee Yam Holding Sdn. Bhd.

    • Individually designed and developed a comprehensive Touch Screen based Point Of Sale (POS) system, Inventory Control System, and Customer Contact Management system using Visual Basic 6.0, ADO, COM, ActiveX Control, and Microsoft SQL Server 7.0.
    • Developed multimedia game systems, including a Horse Racing Game and Roulette Game, using Visual Basic 6.0, DirectX 7.0 SDK, ADO, and Microsoft Access 97.
    • Visual Basic 6.0
    • MS SQL Server 7.0
    • ADO
    • COM
    • ActiveX
    • DirectX SDK
    • POS Systems Development
    • Inventory Control Systems
    • Game Development

Frequently asked questions

What does Lim Chee Kin specialize in?
Lim Chee Kin specializes in enterprise AI and solution architecture. His current focus areas include conversational agents, multimodal document extraction, Retrieval-Augmented Generation (RAG), real-time voice AI, and integration with accounting and business platforms through role-restricted MCP tools.
What is RAG.WTF?
RAG.WTF is an open-source platform for private, personalized knowledge discovery and management using Retrieval-Augmented Generation (RAG) systems.
What is Project Concord?
Project Concord is a Model Context Protocol (MCP) server that connects modern AI agents with legacy database schemas.
How can I contact Lim Chee Kin?
You can reach Lim Chee Kin via email at limcheekin@vobject.com, through LinkedIn at linkedin.com/in/limcheekin, or via GitHub at github.com/limcheekin.
What technologies does Lim Chee Kin work with?
Lim Chee Kin works with Python, TypeScript, Java, FastAPI, React and Next.js for application development. For AI, he uses the Claude Agent SDK, Google Gemini, the Model Context Protocol (MCP), Pipecat for voice, hybrid retrieval, and Ragas for LLM evaluation. His infrastructure stack includes Docker, Kubernetes, AWS, GCP/GKE, PostgreSQL/pgvector, MongoDB Atlas, and SurrealDB.
What is Lim Chee Kin's approach to data sovereignty and private AI?
Lim Chee Kin advocates for data ownership and privacy through self-hosted AI infrastructure. He runs local LLMs, embeddings, and STT/TTS services on dedicated hardware (OrangePi 5 Max with Docker and Coolify), and integrates embedded databases like SurrealDB.wasm for client-side data sovereignty.
What is open-text-embeddings?
open-text-embeddings is a Python library created by Lim Chee Kin that provides an OpenAI-compatible API for diverse open-source sentence transformer models such as BGE and E5. It enables significant cost reduction and eliminates vendor lock-in for text embedding generation in RAG and NLP applications.
Is Lim Chee Kin available for freelance or consulting work?
Lim Chee Kin is open to AI engineering consulting, solution architecture engagements, and collaborations on open-source AI projects. He is particularly interested in RAG systems, local LLM deployment, data sovereignty solutions, and AI agent infrastructure. Reach out via email at limcheekin@vobject.com or LinkedIn.
What is Talk To AI?
Talk To AI is a real-time voice AI application built by Lim Chee Kin that achieves sub-300ms latency for voice interactions. It supports local and cloud-based speech-to-text (STT), large language model (LLM), and text-to-speech (TTS) services via FastRTC, enabling private, low-latency voice conversations with AI.
What is Lim Chee Kin's career background?
Lim Chee Kin started his software career in 1999. He led enterprise Java teams at DXC Technology and Zurich Technology Services, delivered banking and healthcare systems, built 10+ open-source Grails plugins, worked as an independent Principal AI Architect from 2021 to 2025, and joined Daxonet AI as AI Solutions Architect in 2026.