tazi.ai CEO Dr. Tanju Cataltepe gave a talk titled "Concentrating on the Customer with Fast and Accurate Data Analytics" at IBM & SC3 SuperComputer event on June 14, 2017 in Istanbul. The talk emphasized how to speed up tazi.ai's machine learning throughput by means of IBM and NVIDIA's Power8, GPU and NVLink based architecture. First of all, complex problems require multiple machine learning algorithms learning and predicting together. Some of these algorithms, such as deep neural networks, PCA/LDA like matrix operation dependent methods, may benefit from GPU, on the other hand other algorithms, such as decision trees or classifier combiners, may not speedup with GPU. IBM's architecture allows each machine learning building block algorithm to work fast and also allows fast communication of algorithm results on the NVLink bus. Especially with domains such as security, online payment and banking systems and telecommunications, such architectures will bring huge advantages over the existing ones.