See the original article here.

Featured on Google DeepMind's Gemmaverse, this case study highlights our work with SK Telecom to train a multilingual customer service moderation model using Gemma 3 4B. SK Telecom needed a solution that could handle both English and Korean across its 23 million subscribers, while keeping costs and latency low.

Using Adaptive Engine's reinforcement learning capabilities, the fine-tuned Gemma 3 4B model met or exceeded the performance of much larger proprietary and open models at a fraction of the cost, making it a compelling choice for production-scale customer service moderation.