A search engine is a tool that helps to extract our required information with large collections of repositories using data retrieval techniques. As part of related tasks such as searching and data mining, enterprise search engines must be able to analyze a wide range of information resources within the enterprise and employ organizational expertise. Another important group of systems with design goals that differ significantly from commercial search engines are open source search engines. After the initial data collection, it needs to convert these files (code, docstrings) into pairs. These pairs are ideal to collect as training data for code acquisition models. The implementation is Domain-specific optimizations such as B Tree-based P2P can be used to get the most efficiency out of the information in code along with syntax-aware tokenization. After training this model for the frozen version, freeze all layers and continue training the model for a long period of time. This will help us improve the models usage for this work. To quickly find the nearest neighbors, these vectors are inserted into the search index in the final step. To achieve reasonable real-time object recognition, the aim is for at least 10 frames per second. Therefore, Intel NCS cannot reach the 10 fps target. However, it's worth checking out second iteration of the device. This is obviously more powerful and can run arbitrary object detectors in real time. The proposed framework uses semantic and machine learning techniques to extract and model data for specific kinds of IoT applications. Test results show that our system is scalable and can match various IoT applications with a large number of data domains with 80% accuracy

Search Engine, Semantic Search, Domain-Specific Optimization, P2P model, Tokenization, Concrete pooling.