Abstract: “Living matter evades the decay to equilibrium.” Thus said Erwin Schrödinger in his attempt to define life. Microscopic worlds of promise like synthetic DNA nanomachines and metabolic factories in living cells function by actively converting energy and exchanging it with their environments. While the classical laws of thermodynamics paint an exquisite portrait of work, heat, and entropy in macroscopic systems that change very slowly or not at all, a corresponding nonequilibrium theory is lacking completeness. Key unanswered questions are: what principles of nonequilibrium thermodynamics are being exploited by biomolecular soft matter systems? And how can we implement these principles to inform the design of artificial bionanotechnology? Current computational and theoretical methods for nonequilibrium thermodynamic calculations are limited to systems that are either very simple or very “close to” equilibrium. Complex systems evolving very far from equilibrium require a new approach. Here, I present a new method based on automatic differentiation — a technique first developed in the context of neural network training — for investigating nonequilibrium thermodynamics. First, I’ll motivate this technique with applications from my past research investigating biological self-assembly and exploring DNA nanotechnology design. Then, I’ll demonstrate using automatic differentiation to identify how to tune parameters governing nonequilibrium evolution to optimize arbitrary objectives, such as minimizing external work required to drive a nonequilibrium process, maximizing thermodynamic efficiency, or minimizing heat dissipation. Applications range from improving experimental and computational free energy landscapes of biomolecules, to streamlining the design of electronics, to elucidating some of nature’s most astounding molecular machines.