Serverless computing has become a common choice for building scalable and cost-effective applications. It removes the need to manage servers directly and allows functions to run on demand. However, one of the main problems in serverless platforms is the “cold start” issue. When a function runs after being idle for some time, it takes longer to respond. This delay affects user experience and system performance. To reduce this problem, a novel serverless function optimization system is introduced. This proposed system uses a smart prediction model to check when a cold start might happen. Based on this, it prepares the function in advance by warming it up. This way, the function stays ready to respond quickly. The proposed system also improves how resources are used during function execution. It finds the right balance between memory, CPU power, and function duration. This helps in reducing the delay and saving cost at the same time. The method was tested using different workloads, platforms, and function types. The results showed better response times and lower resource waste. Key measurements included function startup time, execution delay, resource usage, and cost. In most cases, the cold start time was cut down by more than half. The system also worked well with both high-traffic and low-traffic functions. This proposed approach helps in building faster and more efficient serverless applications. It supports smooth scaling and better user experience without manual tuning or complex setup.