Adrian Curic

Created a framework that estimates the building's solar rooftop potential, detects buildings from satellite images, rooftop shape, and optimal placement of solar panels, and starts building information profiles using geodata and address matching. Developed a real estate price prediction model that combined hedonic and machine learning approaches and was awarded the WIPO patent, number WO/2022/191775. Built a record-linking framework that allows the consolidation and deduplication of real estate property listings across multiple data sources, which was awarded the WIPO patent, number WO/2022/225455. Developed a framework for address parsing and matching using geodata and natural language processing (NLP). Trained it on over 20 million properties from Singapore, Hong Kong, and Malaysia. Built a floor plan processing framework that extracts property information, generates composite floor plans based on multiple schematics, and restyles the property's floor plan into a design format. Developed a property imputer that processes high-rise building data, identifies valid stacks and residential and commercial floors, and uses imputation techniques to generate data for the missing properties. Built a record-matching library capable of computing similarity for categorical, textual, and numerical features using NLP and unsupervised learning models. Analyzed governmental public transport data to identify daily traffic patterns, compute shop-front traffic numbers, and estimate the impact of property development on commute times. Developed a land lot and address-matching project that can combine real-estate data from unconnected sources based on geodata analysis. The matching can handle complex land lot allocation and unmapped addresses. Built a geodata integration framework to coalesce property data from multiple open source mapping services.

Participated in developing a resource management framework that optimizes the management of medical personnel. Based on intensive care unit (ICU) data, it identifies times when medical attention is most likely to shorten a patient's time in ICU. Researched a model that can predict cardiac arrest events based on real-time data from portable medical devices. Collaborated with Singapore's National University Hospital (NUH) and processed data from ICUs with over two million event record

Contributed to developing surveillance systems based on real-time video processing by working on algorithms for detecting abnormal behavior using flow detection, object detection, and event clustering. Tested and improved system performance for specific client goals. As a result, we won contracts from Singapore's Immigration and Checkpoints Authority and Resorts World Sentosa. Contributed to the deployment of surveillance systems over distributed networks and improved the systems' scalability and robustness.

Served as a consultant and provided technical support for IBM ILOG CPLEX Optimization Studio and IBM ILOG Supply Chain Applications. Developed business model specifications in IBM ILOG CPLEX Optimization Studio that captured clients' business constraints and goals. Collaborated with clients and the development team to identify, investigate, and resolve software issues.

Provided consulting services and technical support for IBM ILOG CPLEX Optimization Studio and IBM ILOG Supply Chain Applications. Identified, investigated, and resolved software issues while working closely with clients and the development team. Documented clients' business limitations and goals by creating business model specifications in IBM ILOG CPLEX Optimization Studio.

Researched the modeling of video stream properties using Markov chain probabilities. Contributed to developing the ASTRA: System-Level Design and Analysis of Architectures for Streaming Applications project. Developed a framework for the simulation of mobile biomonitoring applications. Contributed to the development of the project EASEL: Engineering Architectures and Software for the Embedded Landscape.

Researched methods of partitioning, scheduling, and executing synchronous programs with real-time constraints. Developed tools for code compilation, scheduling, and verification of distributed real-time programs using C, Simulink, and Lustre. Researched methods for scheduling and verifying asynchronous programs in a multithread environment.

Contributed to developing distributed model checking of the LOTOS specification using C and Lotus languages. Developed a library for hash table-based storage methods. Created parallel verification tools, including cluster platforms, hash tables, code generation, and intra-cluster communication.