Case Study: a big tech company fine-tunes GenAI with RWS TrainAI experts

Big Tech Company fine-tunes GenAI with RWS in 4 weeks

Big Tech Company wanted to fine-tune its open-source generative AI large language model (LLM) to maximize its accuracy, safety, and multilingual capabilities. The challenge was that a conventional crowdsourcing approach for data annotation would not suffice; they needed access to highly qualified subject-matter and language experts. They reached out to their existing vendor, RWS, to leverage its TrainAI data services.

RWS's TrainAI team quickly recruited, trained, and managed 285 domain and language experts, employing them as part-time staff to ensure data security. The solution included prompt engineering, model fine-tuning, and red teaming. At its peak, the team completed 10,000–13,000 hours of work per month, providing over 1,200 low-quality examples to improve the model. This allowed the client to successfully train and roll out its latest LLM, making it more accurate and robust, which led to RWS being awarded four additional projects.