Pints AI provides an on-premises plug-and-play AI solution tailored for financial institutions, ensuring the utmost confidentiality and secure handling of data. By merging the strengths of our proprietary large model with industry-specific features, our AI is pre-trained in financial knowledge and seamlessly integrates with our clients' sensitive data. We swiftly empower financial entities, from retail banks to fintech startups, enhancing their sales and operations for a competitive, secure advantage.

Project Role: Next-gen LLM Architecture AI Researcher

This project focuses on addressing the issues within the current transformer architecture, particularly the quadratic problem associated with the attention mechanism. In simpler terms, the current transformer architecture is not optimised, thus making it extremely resource intensive to use.

The solution involves researching alternative AI/ML techniques to replace the attention mechanism with a more efficient one. This unlocks a new realm of LLM capabilities, including the ability to develop smarter AI models, significantly reduce training time, the potential for ultra-fluid/highly adaptive models that can be fine-tuned instantly, and potentially even lay the groundwork for Artificial General Intelligence.

In terms of application, a next-gen LLM architecture could enable compact LLMs to run on standard office hardware or personal devices, unlocking a wide range of use cases and enhancing productivity.

In terms of sustainability, a next-gen LLM architecture would significantly reduce the environmental impact and carbon footprint.

Job Scope:

• Conduct extensive research on possible AI architectures.
• Run experiments and create Proof-of-Concepts to validate findings.
• Collect data and create datasets necessary for model training.
• Develop an AI strategy and pipeline encompassing the necessary tasks and workflow, such as research, data collection, experimentation, and evaluation.
• Assist in writing research papers for publishing.
• Monitor ongoing research within academia or the open-source community for the latest updates or breakthroughs.