The proliferation of various IoT devices equipped with a wide range of sensors, more powerful hardware, along with enhanced machine learning techniques, are all catalysts driving the development of various context-aware systems and applications in both research and industry. More specifically, user mobility and location, have been at the forefront of the contextual information that researchers have been trying to attain due to the large potential of relevant applications. Solutions to date, relying on GPS, rich sensory information, and larger computation capabilities on high-end smartphones, are impractical in many settings where access to these resources are limited. In this work, we propose UniCellular, an overarching platform that will offer more deployable and ubiquitous solutions to obtaining user location and mobility contextual information. UniCellular will only leverage basic single-serving tower cellular information realistically accessible to all devices, by current APIs, and assume minimal device capabilities. We will focus our efforts on the two research thrusts of indoor floor level localization, and outdoor transport mode estimation. We choose these two thrusts because they act as representatives of contexts characterized with different categories of challenges. Thrust 1: Indoor floor level determination. We propose a novel fingerprinting-based ubiquitous and regulatory-compliant floor determination system designed to provide high accuracy using information from only the serving cell tower. The main idea is to fingerprint sequences of consecutive signal measurements to learn relations between the user’s floor and the changes in the serving cell tower associated with the user’s phone and its received signal strength (RSS) value. In this work, We will address challenges related to limited information, oscillating cellular associations and handovers, and various data challenges. Thrust 2: Outdoor transport mode estimation. We will develop accurate, ubiquitous, and easily deployable transportation mode estimation system using only the serving cell tower information. Our system will work on all phones and be available for large-scale deployment. Our solution will have a minimal energy footprint compared to GPS-based solutions. Our efforts in this thrust will deal the challenges of data asynchronicity and unavailability, by training data and creating hybrid learning models that will combine CNNs, LSTMs and MLPs, to integrate automatic feature extraction and time-series processing with feature engineering. We will pursue these thrusts in parallel throughout the duration of the project. In the latter stages of each thrust, we will also investigate integrating edge-based federated learning and user anonymization techniques that will provide more privacy preserving ML-based model building solutions.