Pradeep K C

Working on Development Face Recognition Attendance management system using Deep Learning. Development of face recognition tool using Deep learning tools likes Insightface, MTCNN etc. Using FastApi for the development of api's for face recognition. Conducted extensive research on state-of-the-art deep learning architectures for face recognition, including VGG, ResNet, and EfficientNet, to optimize mode performance. Leveraged transfer learning techniques to adapt pre-trained models for specific face recognition tasks, reducing training time and resource requirements. Collaborated with software engineers to deploy face recognition systems on embedded devices, optimizing performance for real-time applications in constrained environments. Developed a face recognition-based attendance system for educational institutions and corporate organizations, enabling automated attendance tracking and management. Conducted performance optimization and profiling of face recognition algorithms, identifying and resolving bottlenecks to improve inference speed and efficiency. Actively participated in code reviews, design discussions, and technical meetings to ensure high standards of code quality, maintainability, and scalability.

Worked on implementation of machine Learning and Build Platform for Additive Manufacturing Using Python for VSSC, ISRO. Responsible for building and maintaining architecture for the product developed for additive manufactured Alloy. Collection of data from different sources like Api (Thermo-Calc), website and journal paper. Developed the Inhouse SaaS Tool and deployed it using Pyqt5. Continues monitoring of the developed tool and regular interaction with the stakeholders. Regular meeting with the stakeholders on updating of tools and debugging any issue shows up. Exploring and visualizing data to gain an understanding of it, then identifying differences in data distribution that could affect performance when deploying the model in the real world. For predictive analysis, the Machine Learning Tools like Linear regression, SVR, Random Forest Regression, Decision Tree algorithm, and ANN. The dataset contained around 10 features and with the best R-squared 0.95 for train data and 0.89 for test data. The predictive model is stable and not causing any biasing problems.

Worked on Development of Deep Learning Tool for solar, wind and High tension electric Pole's anomaly detection. Converting the Huge Video data into Image data by Using Python script using libraries like OpenCV. Annotation of Images Using Labelme and labelimg. Clearing the blurry, defective images and annotation of different classes for anomalies. Provide analytical insights by analyzing various types of data, including mining and reviewing relevant cases/samples. Investigates data characteristics, complexity, quality and relevance. Through visualizations and summarizations, defines performance, identifies trends and outliers. Image processing Using the OpenCV, PIL Libraries in the Python. Using Tensorflow to Build the Model to predict the Anomalies in the Video data that is captured through Drone. The annotated data is stored in Cloud for Further use. Able to Identify the anomalies at IOU 0.5 for Yolo v5 with accuracy of 70 Percent. The deep learning Models Used Are R-CNN, Faster R-CNN, Yolo-Version like v1,v3,v4 and v5.

Worked with clients in defining clear business problems and with the delivery team in understanding the variables that influenced the customers to migrate to competitors. Performed ML modeling - algorithm XGBOOST, model evaluation and deployment. Created Predictive model for inventory forecasting so that service centers can achieve JIT Standards. Forecast the incident volume in different fields, quarterly and annual. Autotag the tickets with priorities and right departments so the reassigning and related delays can be reduced. Predict RFC and possible failure of ITSM assets.