Function-as-a-Service (FaaS) is a programming model in which developers write event-triggered functions and the FaaS platform automatically manages resource allocation and function execution. FaaS is well-suited for building fog applications deployed at any location on the cloud-edge continuum, delivering both flexibility and resource efficiency. A major limitation of current FaaS platforms is their lack of support for meeting latency and energy consumption requirements. This paper addresses this limitation by exploring the use of Deep Reinforcement Learning, specifically Deep Q-Networks (DQN), to optimize the placement of FaaS functions. The paper proposes FoRLess, a DQN-based scheduler designed to learn the optimal function placement, improving the platform's ability to satisfy latency and energy consumption requirements. This scheduler is integrated into an open-source FaaS platform. The paper describes an experimental evaluation in the Grid'5000 testbed that demonstrates that our approach achieves reductions in latency and energy consumption of up to 7.15% and 12% respectively, compared to the baseline scheduler.