My project uses an artificial neural network to predict the probability of a medical device being compromised or hacked. This project would automate the risk assessment process that medical IT staff use to design a strategy for effectively addressing numerous system vulnerabilities. A need for an efficient and precise risk assessment tool has become more necessary and important with the recent increase in ransomware attacks against hospitals. Some of the biggest weak spots hospital systems have are found in their connected medical devices like EKGs. However, these are not as easy to analyze as software system vulnerabilities and thus take much more time. My project seeks to reduce the time and energy needed to tackle this weakness in hospital systems. 

As the Philippines’ population grows, more solid waste is generated. Poor implementation of segregation policies and the lack of public trash cans have led to solid waste mismanagement in the country. Local landfills are almost at maximum capacity and residents turn to burning garbage as a solution. Unfortunately, majority of these wastes are either made out of plastic or paper which are found to be major pollutants. With this in mind, this project aims to create a budget-friendly trash can that can automatically segregate and shred wastes. This automated trash can and waste segregator is designed to sort the waste into three categories namely; metallic, organic, and plastic. Magnetic sensors are added to determine the type of material, thereby making waste segregation more effective. Additionally, ultrasonic sensors would be placed in all the garbage bins. When the garbage reaches the level of the sensor, then a microcontroller will signal the built-in shredder to operate. 

Recent findings that the COVID-19 case fatality rate (CFR) in the Asian American community is the highest of all ethnic groups in the U.S., and that COVID-19 CFR is not higher in Asians in other parts of the world, points toward the impact of health disparities. This has motivated a project at the University of Louisville to develop a GIS-enabled software to provide peer-to-peer support to help the first and second generation Asian American population navigating the complex healthcare system. The Asian Institute Crane House in Louisville, KY is the community partner for the project and Phase I of the project December 2021-September 2022. Community AI’s founder Shreas Kar was part of the research team in the prototyping stage of the mobile app development.

CharityTrend is a website that is used to recommend charitable organizations based on news stories. Users paste an article link into CharityTrend‚ search bar, and an AI model will analyze the article‚ content for keywords that reveal the issues that are discussed. The model will then search for relevant charities. The user can also view trending news stories with charities that the model has picked. CharityTrend will also display additional information about each charity so that users can evaluate them further. Afterwards, users will be navigated to the donor page to make a secure donation.

Incentiva is a minimalistic task manager that is packed with features to help significantly improve the productivity of the user. Incentiva aims to keep users productive and going in their busy day-to-day lives, ‚incentivizing‚ users to keep their good habits and to drop their bad ones. Advanced AI analyzes the user‚ habits to show them detailed reports on what they should be doing more/less of. Incentiva is different from other task managers in that it actually ‚manages‚ your tasks. It motivates users to keep going, and includes productivity features such as a Pomodoro Clock to ensure their work is getting done in a timely manner.

Global tree cover loss rose from an average of 17.1 million hectares a year in the 2000s to 23.1 million in the 2010s. The loss of forests has several detrimental effects on the environment such as the acceleration of global warming, the increase of biodiversity loss, the disruption of rainfall patterns, and many more. This project focussed on three solutions to combat deforestation: reforestation, eco-forestry, and green agriculture. All three solutions will use a combination of neural networks, image processing, satellite imagery, and classification.

With institutions open across the country during this pandemic, we are increasing the risk of outbreaks. On-location activity is particularly important for traders who need robust, real-time communication and sales teams that are subject to specific compliance monitoring. In the event of an outbreak, an institution may not be able to accurately identify every employee/student at risk and in turn, can be forced to shut down completely. There is a huge opportunity cost due to a loss of learning potential/revenue from shutdowns. OPEX is a startup focusing on providing solutions for contact tracing. Our long-term vision is to provide asset tracking solutions maximizing the efficiency of manufacturing infrastructure.

In the United States, and especially in urban areas, child abduction is a serious issue. When this happens, an amber alert is sent but this is largely ineffective and drains police resources. This project automates the vehicle detection process by constructing an AI in traffic cameras to detect and identify similar vehicles to that of the child abductor. This is done through frame differencing to find the vehicles in the video and then a convolutional neural network to compare features of the vehicles. As a result, criminals in urban areas can be found more easily, quickly, and safely. This will help discourage criminal activity and create a safer environment for children.

In Boston, the budget of the MBTA transportation system continues to increase, yet the majority of citizens are unsatisfied with its service. In recent decades, inequitable transportation has received increased attention on its impacts on economic disparity and families, costing industries millions. This project aims to make bus transportation more efficient, reliable while reducing its costs on the MBTA’s limited resources. Using a polynomial regression model to predict the number of people who arrive at the bus stop every day with a number of different variables including time of day, area, and day of the week, this accurate prediction enables much greater efficiency in creating bus schedule timings. An algorithm used on this regression curve, a function of the number of people arrived at the stop will optimize departures, evening them based on how many people are at the station. This prototype could then also provide very useful data its predictive power, to also be able to numerically predict the number of people currently at a bus stop.

Everyone around the globe is all too familiar with the term ‘coronavirus’. As this disease plagued our lives for the worst and took away too many others. At the end of the tunnel we saw light, the Approved usages of vaccines that have been clinically proven to combat covid. However, due to mass misinformation circling the internet, it’s not surprising that many are hesitant to receive the vaccine. Not all is lost, with the assistance of a machine learning app. Through the use of neural networks and multivariable linear regression, it will analyze the user’s age, health conditions, etc. to recommend the user correlating articles with reliable sources from doctors and researchers. It can show the individual nearby vaccine locations and what vaccine(s) they offer. This app will serves as a guide or rather a helping hand to educate someone regarding the importance of getting vaccinated and make booking appointment(s) less of a hassle. At the end of the day the purpose of CAYF is to nudge everyone to receive both doses of the vaccine to achieve herd immunity so no one has to live in fear of a threat that cannot be seen by the human eye. 

Landslides cost billions in damage annually in the U.S. & have affected 4.8 million people over the past 2 decades. Existing landslide warning systems are ineffective in accuracy, latency, & scalability. There is a lack of publicly accessible data of landslide features on a global scale. The researchers compiled a global landslide dataset consisting of ~18,000 landslide incidences and 100 related climate and terrain features. Machine learning models obtained 92.5% accuracy & 94.1% detection rates when forecasting whether or not landslides would occur 5 days into the future. A data-driven approach to landslide susceptibility map generation was derived.

Mental health (emotional, psychological and social well-being) affects how we think, feel, act and therefore, our response to stress and the behaviour of people surrounding us. It decides whether we make healthy choices or not. However, various factors such as cultural biases, uneducated assumptions and very similar symptoms of more than one type of mental illness can lead to misdiagnosis of diseases. Studies show 2 out of 5 patients with bipolar disorder are always misdiagnosed with major depressive orders, at first. The mean period of delay in bipolar diagnosis ranges from around 6 to 8 years. This project aims to utilize artificial intelligence to minimize (if not completely eradicate) this problem of misdiagnosis. It plans to make computers learn the symptoms and characteristics of each illness, their similarities with other illnesses, the limitations of current diagnostic tests and many other important factors to give a proper diagnostic report of a certain patient. It aims to consider all sorts of medical data available to provide the report, starting from the digitalized data of the blood samples of the patients to results of various conventional diagnostic tests undertaken by the same set of people.

Machine Learning Algorithm (Convolutional Neural Networks) to Detect Icebergs from Satellite Images to Help Reduce Global Warming, Oil Spill and Improve Safety of Arctic navigation

By Shraman Kar

By: Amy Chen and Kaavya Thirumurugan

Just the other day I read a headline, farmer couple commits suicide after killing minor‚ and this made my heart skip a beat. And I thought to address the problems of yield losses due to diseases with what I love, tech. Plant AI is a web application that helps crop growers to easily diagnose diseases in plants from plant images, using Machine Learning possible all on the web. This relies on a model I built to identify plant diseases from images, that runs on-device to prematurely identify plant diseases and also suggest farmers actionable steps to solve it.