The project involves non-invasive bilirubin estimation in humans with a Near-Infrared (NIR) sensor and machine learning to provide a painless and safer alternative than regular blood tests. Various blood types will reflect and absorb light differently so that the NIR sensor can take measurements of spectral data common in the subject’s bilirubin level. This information is transmitted to a laptop where it is processed through an LSTM (Long Short-Term Memory) model, a form of recurrent neural network, and categorized as one of the high bilirubin levels (Normal, Moderate, Critical). The model is trained on existing, labeled NIR spectral data to become very precise, even for real-time prediction. The system described provides a contactless, sanitary process, particularly well-suited for situations of high risk or distant clinics where standard laboratory facilities are not easily accessible. It is hoped that this project will improve the speed, safety, and ease of blood type determination through the combination of NIR spectroscopy and deep learning. The combination of embedded hardware and AI provides a modular and scalable architecture that can be easily deployed in both home-care and clinical environments. The contactless operation of the system provides a hygienic and painless experience for the subject, reducing cross-contamination risks. Utilization of low-cost elements also adds to accessibility, especially in resource-scarce areas.