LOCUST is a low cost intelligent transportation system that uses smartphones to create a virtual road orchestration network.
• Worked on the detection algorithms for potholes, traffic jams and driver behavior, using the Hadoop ecosystem
• Designed the machine learning algorithm that runs on mobile devices to increase the accuracy of the data sent to the platform and help retain the user battery
• Designed the routing engine based on genetic algorithms
Big Data, Machine Learning, Genetic Algorithms, Hadoop Ecosystem (HDFS, MapReduce, pig, Hive ,flume ,kafka ,Hbase, sqoop, storm)
Create a toolbox to analyze grey matter, across two different populations (healthy subject and controls) and generate meaningful network patterns and brain structural insights.
• Joined a multidisciplinary team working on the brain analysis project, required adapting to a new field with different taxonomies and tools.
• Analyzed the best implementation for a brain analysis toolbox, researching relevant APIs and different analysis techniques.
• Used the information I collected, and developed toolbox that extracts viable insights from brain connectivity networks.
Computational Neuroscience, Machine Learning, Matlab, Python, R, Weka, Image Processing,
Graph Theory, Brain Connectivity Toolbox.
A natural language processing (NLP), and a machine learning with the purpose of extracting keywords from long narratives to be later used in text clustering as well as information retrieval systems.
• I implemented and analyzed RAKE to determine the reasons behind its drop in performance when applied to long narratives.
• I examined other keywords extraction approaches, to get more insight into NLP best practices, since this was my first project in the field.
• I finally, developed my own approach, that remedied the shortcomings of RAKE, and improved its performance as evident by the increase of F-Measure from 5.72 to 32.2
NLP, Machine Learning, RAKE, Java, Python, R, Weka, Stanford NLP Toolbox
The purpose of this project was to build an observation layer on top of the RoboCup Rescue Simulator, to observe the behavior of the various agents, detect flaws, and ultimately improve the performance of the system, which is used to simulate natural disasters and
• Configured the RoboCup Rescue Simulator
• Created a an observation layer for each type of agent
• Intercepted and logged actions and communications between the different agents
• Analyzed the observation logs in tandem with the simulation logs to detects erroneous patterns and improve the default behavior of agents
RoboCup Rescue, Artificial Intelligence, Multi Agent Systems, Parallel Programming, Java, ant