Using Deep Learning to Find the Next Unicorn: A Practical Synthesis