Himanshu Sharma Email & Phone Number

Indore, Mp, In

Bengaluru, Karnataka, India

1. Developed a book recommendation system for amazon customers using memory and model-based collaborative filtering by utilizing the book description and user interests.2.Used Word2Vec vectorizer on descriptions to obtain user-item interactions and employed cosine similarity to measure user-item and item-item similarities.3.Created profiles for top active users by leveraging interaction strength with the recommended items and achieved test recall@5 of 42% and recall@10 of 53%.4.Handled the cold start problem based on global and demographic-specific book popularity and improved the efficiency of the user recommendation engine by 33% .

1. Developed multi-class XGBoost model to characterize the email creative and predicted its effectiveness that is ignored; read; acknowledged by the reader.2. Performed missing value imputation using KNN-Imputer, implemented SMOTE boosting, and carried out hyperparameter tuning using Bayesian Optimization.3. Leveraged SHAP plots to identify key model drivers such as limit of word count, keywords, communication time, and personalization.4. Obtained F1 scores of 89% & 84% on train and test data respectively and estimated an overall increase in the customer acquisition rate by 15%.

1. Built an XGBoost regression model to predict the popularity of news on different social media platforms such as Facebook, Linked In, and Google Plus.2. Performed key NLP processing steps on news items such as lemmatization, POS-tagging, Named Entity Recognition, stopword removal, and tokenization using spaCy.3. Generated the optimal document-word matrix as model input by tuning parameters such as min_df, max_df, n_gram range, token_pattern in TF-IDF vectorizer.4. Performed hyperparameter tuning using GridsearchCV & RandomSearchCV and achieved R2 score of 72%, 76%, 78% respectively for FB, LinkedIn & GooglePlus.

1. Built an IoT-enabled Web app using Python, Flask, and Sqlite where students can register themself and faculties can sign in and initialize the attendance process.2. Encoded a picture using the HOG algorithm to create a simplified version of the image. Found the part of the image that looks like a generic HOG encoding of a face.3. Passed the centered face image through a Deep Convolutional Neural Network that knows how to measure features of the face and saved those 128 measurements.4. Developed an SVM classifier to find the person in our database of known people who have the closest measurements to our test image.5. Obtained an accuracy of 99.8% and able to deliver an accurate face recognition attendance system with minimal human intervention, less time-consuming compared to manual attendance system, and reduced discrepancies like proxy attendance .

Bachelor Of Technology - Btech, Computer Science, 9.24 Cgpa

Activities and Societies: Table tennis, Football, Cricket

Higher Secondary, 67.0%

Secondary, 7.0 Cgpa