Final projects for the real world
Rather than test students on what they've learned in the traditional sense, the MITS program pairs them with research centers or companies to solve a real-world problem. Learn more about the 2022 Capstone Projects and watch videos of the presentations at the links below.
Business Intelligence Tool Development (ProCogia)
Team members: Cristin Connerney, Ian Baird, Chenjun Zhou, Jiarui Zhang
ProCogia is the home of data science, supporting our clients at all stages of their data journey. From ETL, data preparation and Data Warehousing right through to Advanced Analytics, Data Science, Machine Learning and Bioinformatics.
ProCogia is developing a vertically integrated in-house tool to streamline the management of its day-to-day operations, members of this project will act as consultants to the ProCogia team aiding them during the development of this tool.
Tweet-O-Meter: A Probabilistic Classifier For Disinformation
Team members: Prakhar Agrawal, Manisha Dasiah, Junyi Guo, Xinran Ma
The basic need of this project is to be able to identify those messages on social media that are spreading disinformation, as such messages compromise business practices, are used to recruit users to products or groups that are harmful or scams, and are used to create fear. Having a working system for tagging messages with such contact would help address these needs.
The cost of hand-tagging messages as containing disinformation is prohibitive. Human-based tagging creates severe burnout in the humans, and can depending on the content, cause mental unrest or even trauma in the human coder. An automated system, or even semi-automated would reduce these harms.
The Great World Powers, Twitter, and Ukraine
Team members: Natalie Holley, Haoyu Niu, Jun Otani, Anthony Perry
Data will be collected from at least three social media channels – Twitter, Facebook, and YouTube.
This data will be collected around the state sponsored media and the diplomatic, executive and foreign ministers for each country. The countries of focus are China and Russia, and the receivers of their social-cyber communiques – i.e. the US, neighbors, etc. This set of countries needs to be discussed and finalized early on.
This data will be coded for diplomatic activity and whom it is directed at and the nature of that activity. This data will also be assessed from an influence campaign perspective using the BEND maneuvers. The team will be taught these maneuvers.
The bulk of the project is synthesis and analysis. Understanding what is being said and why. It will involve contrasting this social cyber face with other ongoing activities and historic lines of diplomatic effort by these countries. It will involve comparing the approaches of China and Russia to their allies, to another great power (US), to regions that could be annexed (e.g. Taiwan and Crimea), and to their “adversaries.”
Oh!Lab Twitch Toolkit
Team members: Jeremy Chen, Ruijie Huang, Yi Sun, Ran You
To date we have created three separate systems to facilitate working with game streaming in different capacities. (1) A toolkit for audience participation games, which primarily uses IRC chat to enable stream viewers to interact with a streamer’s game. (2) A toolkit for game-aware streams, which leverage meta-data logged from inside a game to be referenced by a Twitch extension on a viewer’s machine to augment their viewing experience. (3) A researcher toolkit, which provides researchers with tools to capture streamed video, viewer chat, and other related signals from game streaming services into a central place that can be temporally aligned across viewers. The primary goal of this project would be to develop some new features for the researcher toolkit (system 3), however a secondary goal would be to provide strategic help in unifying the systems into a common architecture that could be better positioned to attract additional engagement from external collaborators or industry partners.
TCS Automated Driving Project
Team members: Manish Agnihotri, Joshua Kim, Yumeng Li, Yinghuan Zhang
The objective is to detect all unique objects available in an image captured by test vehicle under Autonomous Driving Assistant Program. This shall help in building a glossary of all objects of interest to be used under verification and validation processes and expedite the process by just not only relying on known objects (pre-trained models). Using this approach, training of models for different objects of interest can be avoided
Automated Threat Modeling Analysis
Team members: Henry Fox, Nathan McFadden, Shankara Narayana, Addison Whitney
Planning and executing Threat Modelling (TM) is very crucial to address security vulnerabilities to develop resilient system. However, it is very challenging to integrate threat modelling in modern SW development practices like DevOps. It is usually manual, and siloed activities. To solve this problem, threat modelling should be integrated into DevOps pipeline as developing machine-readable format of the developed model and assess, execute and maintain the model during continuous integration and delivery/deployment practices.