Road anomalies such as potholes and speed bumps pose significant challenges to driver safety and passenger comfort. Despite modern vehicles increasingly employing sophisticated sensors to mitigate such hazards, existing frameworks and infrastructural constraints make these solutions exceedingly expensive. This paper suggests a low-cost, small, and modular system that could be deployed in smart cars, uses open-source tools and readily available hardware to detect road irregularities in real-time, and adjust the vehicle's speed appropriately. The core of the system is a custom-trained, lightweight YOLOv8m object detection model to recognize road anomalies with high accuracy and optimized to run inference on limited hardware. The overall framework comprises two intertwined nodes: Raspberry Pi acts as the vehicle's hardware controller, managing GPIO-driven dual motor control, allowing for smooth speed control, and hosting a Flask-based streaming server. Meanwhile, a remote pc connected to the same network runs a detection script that takes the live video stream of the Raspberry Pi, detects objects within every single frame, and sends real-time control commands using the MQTT message protocol. The system GUI also provides a visual interface for the user to manually start or stop the vehicle. To ensure smooth and reliable operation, buffered detection logic, confidence threshold, speed ramping, and falsification mechanism were also added to the speed controller system. Experimental results demonstrate that the system performs well in real-world situations. The performance was evaluated based on detection accuracy, inference time, latency, and responsiveness of the speed control module.