We present an artificial intelligence-based algorithm to efficiently reduce the time required in the design of multipump Raman amplifiers operating simultaneously in conventional (C) and long (L) bands. The performance of the amplifiers was measured in terms of on-off gain, ripple, optical signal-to-noise ratio, and noise figure, considering a single-mode fiber (SMF), a dispersion compensating fiber, as well as a photonic crystal fiber. Beyond the time reduction provided by extreme learning machine (ELM) and particle swarm optimization (PSO), the numerical simulation results show optimal gains for all fibers in the C  +  L band. A comparison between the proposed algorithm, the standard PSO, and the budget heuristics + multiobjective optimization based on a nondominated sorting genetic algorithm was performed. The forecast established by the ELM in the three fibers specified a root mean square error of 0.0195 in the pump wavelengths and powers test set, with a computational time of 52 s. The simulation results of the proposed PSO-based multiobjective optimization with four pumps after 100 km of SMF demonstrate an on-off gain of ∼0.5  dB higher, when compared to the above-mentioned two methods.