Today's hot topics, data science and machine learning, offer many algorithms that can turn lifeless data into useful information. Based on math-heavy theory, relying on number crunching, often billions and billions of raw operations must be performed quickly. Clojure can be teamed up with hardware accelerators or GPUs to offer both a cozy dynamic REPL environment and the speed and power of low-level hardware optimizations for numerical computations. In this talk, I present a few libraries tailored by the Clojure's measure, that can help achieve state-of-the-art performance: ClojureCL to direct the GPU, Neanderthal to help with vectorization and linear algebra, and Bayadera's incredibly fast MCMC engine. Clojure may not be able to compete with R and Python in the off-the-shelf machine learning, but for high-performance customized algorithms - it can be THE secret sauce.