Accurate prediction of the solar wind speed and arrival of solar transient events are among the most important but still open problems in space weather research. This article implements a Convolutional Neural Network (CNN) coupled Long-Short Term Memory cell (LSTM) deep-learning model for the prediction of solar wind speed at Sun-Earth L1 on a daily and 6-hourly basis. The model is trained and validated using Solar-wind speed observations at Sun-Earth L1. The proposed CNN-LSTM models are developed and trained using data for the period from 01 August 1996 to 31 December 2019 and tested with the data from 01 January 2020 to 30 June 2024. The daily prediction model can forecast solar wind speed with a RMSE of 52.12 km/s and an overall correlation coefficient of 0.79, while the 6-hourly model can forecast solar wind speed with a RMSE of 33.72 km/s and an overall correlation coefficient of 0.93. This shows that without explicitly building in physics based relationships, the Deep Learning based models are able to perform reasonably well compared to existing approaches.