In biotechnological processes, biological process parameters such as biomass concentration, nutrient concentration, chlorophyll content, and product quality can be challenging to measure online. Typically, these parameters are measured in laboratory settings employing offline sample analyzers. Yet, offline measurements cannot be used as quick feedback signals for process control because of the significant delay between sampling and result generation. Moreover, though generally very accurate, they are equally expensive and often come with high maintenance costs. Therefore, soft sensors are widely utilized to address this problem, allowing reliable online estimations of these essential biological process parameters. Deep learning-based soft sensors are prevalent these days due to higher prediction performance. This work describes the use of Long Short-Term Memory (LSTM), a deep-learning architecture specifically designed for time series data, to predict microalgal growth. LSTM has the advantage of predicting future values based on history. The LSTM-based soft sensor developed for predicting microalgae biomass shows higher prediction performance than the support vector regression (SVR) based soft sensor. An LSTM was trained on the indoor cultivation data of Nannochloropsis cultivated in a vertical flat panel photobioreactor. The dataset consists of 28,741 samples, with 20,280 used for training and 8461 used for evaluation. Our LSTM-based soft sensor performed better than the SVR-based sensor and achieved an R² score of 0.91, which was higher than the R² score of 0.781 achieved using the SVR-based soft sensor.