<p>Poor handwriting of doctors in medical prescriptions can cause misunderstanding, misreading, misinterpretation risks and cause medical errors. Despite being rare, identifying the correct drugs with the unclear prescription seriously affects patients and also requires a quite lot of time, attention, and effort for recognition. As a result, a drug name recognition system is implemented, but previously developed models are not significant in terms of accuracy, interpretability, and reliability. Therefore, the skill split optimization enabled large language-based time-distributed bidirectional long short-term memory (SkSpO-L2TDBM) model for drug name recognition is proposed in this research. The SkSpO-L2TDBM model exploits deep features concerning bidirectional encoder representations from transformers and bidirectional long short-term memory techniques that are employed to increase the model’s reliability and interpretability for effective recognition. Moreover, the SkSpO algorithm tunes the hyper-parameters of the proposed model based on the effect of skillful learning and sharing ability that makes easy recognition with maximum convergence speed. The major advantages of the proposed model are simplicity, robustness, and endue complex computation for accurate recognition. Compared to other existing techniques, the SkSpO-L2TDBM model achieved a minimal mean squared error rate of 4.46, and minimal root mean squared error rate of 2.11 using the hybrid dataset comprising the ‘Handwritten Medical Prescriptions Collection’ dataset, and a proprietary set of handwritten medical prescriptions collected from various doctors across the cities such as Nagpur, Pune in Maharashtra, India. Moreover, the proposed approach is robust in recognizing the biomedical entities in the text.</p>