Abstract: Chemical reaction networks (CRNs) taken under mass-action kinetics play a central role in the mathematical modeling of chemistry and biology. A key reason for their widespread utility is their capacity to exhibit multiple attractors and capture a wide range of nonlinear phenomena. Computing paths of transition between attractors, or instantons, is a challenging task, not solvable analytically for all but the simplest cases. In our work, we propose an algorithm for numerically estimating instantons for a CRN. The algorithm uses the Hamiltonian description of a stochastic CRN and solves a MinMax problem on the Action functional to converge on the instanton. In this talk, I will present a schematic derivation of the Hamiltonian and Action functional for stochastic CRNs, explain our Action Functional Gradient Descent (AFGD) algorithm, and show computational and practical applications. Finally, I will briefly discuss the unified formalism to which both stochastic and quantum Hamiltonians belong and propose directions for future research. (For details, see