TELUS International Launches Experts Engine to Revolutionize GenAI Model Building

TELUS International has unveiled Experts Engine, a fully managed solution for sourcing tech-enabled experts-on-demand specifically designed for generative AI (GenAI) model builders. This platform pairs human expertise with GenAI tasks, such as data creation, collection, annotation, and validation. The aim is to expedite, enhance accuracy, and ensure cost-effectiveness in generating high-quality training datasets at scale, particularly for demanding foundation models like LLMs.

Experts Engine is an integral part of TELUS International's comprehensive AI Data Platform, which includes task execution platforms and tools for project and people management. Experts Engine relies on a proprietary machine learning (ML) model that thoroughly assesses the domain expertise, proficiency, education level, and availability of over a million contributors within the company's diverse global AI Community.

 “Training data is paramount in defining the quality, alignment and safety of models, which is why we have enhanced our homegrown AI Data Platform with Experts Engine to address more complex and customized dataset requirements. With today’s launch of Experts Engine, we are taking another significant step forward in our ability to solve key challenges that clients often struggle with when fine-tuning their models at scale by seamlessly combining automation for speed and efficiency and human perspective for quality and reliability. This ultimately mitigates bias and guides models toward safe and inclusive content for use cases spanning all applications and industries,” said Siddharth Mall, Vice president, Product for TELUS International AI Data Solutions.

Tailored to project requirements  

It also evaluates the subject matter, complexity, and sensitivity of GenAI tasks and task types. Tailoring its approach to project requirements, the model matches tasks to the most qualified individuals and assigns appropriate quality control workflows. This strategic approach eliminates the over-allocation of expertise to common tasks, ensuring that the right contributors handle the right tasks efficiently at all times, particularly in large-scale data projects.

TELUS International's Experts Engine solution draws its strength from a diverse global AI Community of contributors, forming the core of its architectural framework. This community plays a crucial role in handling diverse data types across 500+ languages and dialects. With expertise spanning various fields and task types, these contributors represent a broad spectrum of demographics, geographies, education levels, and cultural backgrounds. This diversity is strategically cultivated to meet stringent requirements for GenAI model alignment and safety, incorporating processes such as prompt engineering, supervised fine-tuning (SFT), reinforcement learning from human feedback (RLHF), and other human-in-the-loop methodologies.

The company's expert pool is designed to efficiently manage large volumes and a multitude of task types, by offering flexible workforce models, including community-based, in-facility, and hybrid on-demand approaches. TELUS International is actively involved in engaging, managing, and monitoring members of its AI Community, ensuring their readiness to tackle new tasks.

Elsewhere, TELUS International has partnered with Five9 to introduce an upgraded version of TELUS's Contact Center as a Service (CCaaS) application platform.