In my project, I used a similarity-based approach with the Cosine Similarity algorithm to recommend food items based on user preferences and nutritional requirements. Both user inputs and food items were represented as numerical feature vectors consisting of nutritional attributes such as calories, protein, carbohydrates, and fats. Instead of using a complex predictive model, this approach focuses on comparing nutritional profiles to find the most relevant matches. Cosine similarity measures how similar two vectors are by calculating the cosine of the angle between them, which makes it suitable for recommendation systems where relative feature importance matters more than absolute values. For each user input, I calculated cosine similarity scores with all available food items and ranked them to generate recommendations. I chose this method because it is efficient, easy to implement, and works well with high-dimensional nutritional data without requiring a large labeled dataset or extensive training.