IRAI Labs - Independent AI Research Organization

Hong Kong AI Research & Development

Independent AI research, custom models, and coffee consultations for Hong Kong's growing AI ecosystem

Custom AI Models

We build custom AI models for organisations that need more than off-the-shelf solutions. Every engagement is grounded in rigorous research methodology, with an emphasis on explainability and reproducibility.

Our work spans foundational model development, domain-specific fine-tuning, and specialised AI systems. Engagements range from targeted technical advisory to multi-month research collaborations.

Inquire About Custom AI

Coffee for AI Advice

We trade AI advice for coffee. One of our core missions is to strengthen Hong Kong's AI ecosystem, and making expertise accessible is the most direct way to do that. If you run an SME or startup in Hong Kong and want to talk through your AI challenges, buy us a coffee and we will sit down with you - no pitch, no catch, just honest technical advice.

These sessions are informal and obligation-free. We can help you think through:

  • Whether AI is the right approach for your problem
  • What data you would need and whether you have it
  • Which off-the-shelf tools might already solve your problem
  • Where to start if you decide to build something custom

If you are building a product, optimising operations, or simply exploring what AI can do for your business, we are happy to talk.

Buy Us a Coffee

Open Call for Research Grants

IRAI Labs provides compute resources and mentorship to researchers pursuing ambitious ideas in machine learning and artificial intelligence. Our grants are open to independent researchers, graduate students, and anyone with a compelling research direction who lacks access to the computational resources needed to explore it.

We evaluate proposals on scientific merit, feasibility, and potential for impact - not commercial viability. If your work advances fundamental understanding or opens new research directions, we want to hear from you.

Submit a Proposal

Publications

  • LLM-ADE: Large Language Models with Adaptive Data Engineering

    Stephen Choi, William Gazeley • 2024 • arXiv

    Our knowledge injection framework, which uses selective block freezing and expansion to adapt to new data without catastrophic forgetting.

    LLM-ADE diagram illustrating adaptive data engineering
  • ConFIRM: Conversational Factor Information Retrieval Model

    Stephen Choi, William Gazeley, Siu Ho Wong, Tingting Li • 2024 • ICAIF 2024 Workshop

    Generates domain data using OCEAN personality traits and style transfer to fine‑tune LLMs for high‑accuracy, domain‑specific retrieval.

Supporter

We are an active supporter of AI Tinkerers Hong Kong, a community of makers, researchers, and enthusiasts exploring the frontiers of artificial intelligence.

They are the only technical AI community in Hong Kong, and hold monthly meetups where community members share their projects and ideas - if you are an AI practitioner, you should join them.

Contact