An embedded system is a system consisting of software code, controller hardware, and Input/Output hardware that performs a specific task. There are several challenges when developing an embedded system. First, the code often involves hardware configurations that require domain-specific knowledge. Second, the hardware may have code usage patterns that should be followed. To overcome such challenges, we propose a framework called ArduinoProg towards automatic generation of Arduino applications. ArduinoProg takes natural language queries as input, then outputs hardware configurations and code usage patterns of the hardware for the query. Motivated by our findings on the characteristics of real-world queries posted in the official Arduino forum, we formulate ArduinoProg as three components, i.e., Library Retriever, Hardware Classifier, and API Generator. First, Library Retriever preprocesses the input query and retrieves a set of relevant library names using either lexical matching or vector-based similarity. Second, given Library Retriever’s output, Hardware Classifier infers the hardware configuration by classifying the method definitions from the implementation files of a library into certain communication protocol classes. Third, API Generator leverages a sequence-to-sequence model to generate the code usage patterns also based on the Library Retriever’s output. Having instantiated each component of ArduinoProg with various machine learning models, we have evaluated ArduinoProg on real-world queries. The performance of Library Retriever ranges from 44.0%-97.1% in terms of Precision@K; the Hardware Classifier classifier can achieve 0.79-0.92 in terms of the area under the Receiver Operating Characteristics curve (AUC); API Generator can yield 0.45-0.73 in terms of Normalized Discounted Cumulative Gain (NDCG)@K. Demo: https://youtu.be/d8E4Zjrs_KQ