Merlion AI

Merlion AI is an innovative Singapore AI R&D group with a vision to revolutionize industries worldwide through the power of artificial intelligence and to establish Singapore as an Asian and global leader in generative AI. Our mission is to develop state-of-the-art generative AI models that enable businesses to build effective in-house solutions.

The founding team comprises experienced entrepreneurs, influential public leaders, and world-class researchers and software engineers, including former developers from Google DeepMind and Meta. Our developers are exceptional, belonging to an elite cohort of professionals who have contributed to well-known NLP models.

With our experience, knowledge, and technology know-how, we can build models with training efficiency at least 10 times higher than current LLMs, resulting in the highest value/cost ratio, competitive prices for users and high ROI for our investors.

To ensure high-quality datasets for training our models, our early investors have deep connections with various content providers. Combined with our unique retrieval-augmented, RLHF and self-alignment techniques, our models will provide unprecedented levels of safety, accuracy, and alignment.

We prioritize privacy guarantees and implement strict confidentiality protocols and robust security measures to safeguard sensitive information. Our focus is on deploying models on private clouds or devices, offering an additional layer of control, security, and minimizing external dependencies and vulnerabilities.

A closed technology approach where the models are kept secret and only accessible through APIs raises concerns regarding data privacy, model safety, and integration with other components. Businesses no longer want to feed sensitive data into a “black box” system. Merlion AI will utilize an оpen-source strategy for its foundation models and build a vibrant developer community providing support for developers and researchers. This will boost the community members and their networks' R&D and downstream application development. Openness and transparency will promote effective integration. Different customer content components can be fed into the corresponding parts of the models designed to handle that specific data. However, the most advanced and specialized models will be non-open-source to allow the business to generate enough cash flow.

Our business strategy is based on model integration and partnerships with industry leaders, integrators, and companies developing Gen-AI solutions. We also focus on commercializing specialized models, including the best, latest, and most powerful models, as well as models of small capacities for SMEs.

Merlion AI’s Three-Stage Roadmap includes:

1. Building a family of competitive foundation models in the first year.
2. Developing specialized models with plugins for domain-specific purposes and multimodal inputs in the second and third years. This includes an open-source text-generative model with text and visual output, as well as different versions of small but helpful AI assistants that can run on regular laptops. We will continuously enrich training datasets through partnerships and data acquisition.
3. By the fourth year, Merlion AI aims to become an Asian and global research leader in AI, trusted by major public and private institutions for its safe and efficient technology.

To achieve our goals, we are assembling a world-class team of researchers and software engineers specializing in deep learning, reinforcement learning, natural language processing, computer vision, and data science. If you are interested in joining us, please email your resume and a brief introduction of your expertise to info@merlion.ai.