Hinglish has some problems in sentiment analysis in terms of code-switching, informal writing, and transliteration. Such linguistic complexity is usually out of the scope of conventional NLP solutions as they employ rules and heuristics, especially about user-generated content such as comments to videos. This work proposed the usage of deep learning models for the code-mixed video comment sentiment analysis on Hinglish dataset. We, therefore, consider multiple deep learning approaches with a focus on the Bidirectional Long Short-Term Memory (BiLSTM) that is enhanced by the Gated Recurrent Units (GRU) to analyze the mixed-coded texts accurately. With a test accuracy of 91.7%, our model can capacitate in cases with problems created by code-switching and colloquialism. The results, reveal to what extent BiLSTM with GRU can capture the changes in context and semantics’ cohesion in the Hinglish video comments. This research has implications for increasing analysis of consumer engagement and/or sentiment detection in the faster-growing social media networks. It also gives some ideas on the possible application of deep learning methods for sentiment analysis in a code-mixed environment.