Video streaming traffic has gone through an impressive growth over the last ten years. This growth can be rooted back to two different causes, namely widespread content accessibility and improved delivery techniques, such as HTTP Adaptive Streaming (HAS) and Web Real-Time Communication (WebRTC). HAS techniques are widely used in Video-on-Demand and live streaming scenarios, while WebRTC is mostly used in interactive streaming applications. Both HAS and WebRTC are unfortunately still affected by several inefficiencies. In HAS, the video client is equipped with a heuristic to dynamically adapt the video quality to the available network resources. This client-based adaptation cannot always guarantee the best experience to the end users, the so-called QoE. On the other hand, WebRTC is by design peer-to-peer, which makes this technique not scalable when the number of users in the system is high. To solve these problems, this PhD thesis proposes an advanced architecture where additional intelligent components are placed in the network to support the delivery of the video. Moreover, the designed network components aim at optimizing specific QoE parameters that directly impact the users' viewing experience, rather than QoS parameters, which represent low-level network performance. This architecture has been applied, for example, to reduce playout interruptions in the video playback of HAS clients up to 45% or to increase the scalability of WebRTC systems without affecting the delivered video quality.