Boeing SCDP Project Management

Due to the confidentiality of this project, details on the technologies tested and the solutions realized have been omitted.

Summary

In facility security efforts, it is common to put a few guards in front of hundreds of screens. One source admitted their officers only use 2.5% of the camera videos available to them. While this is sufficient for their needs, 97.5% of the cameras are statistically more prone to rare events. My team created a solution that simplifies the job of security professionals by using machine learning to highlight anomalies in a virtual environment which naturally maps information onto the surveilled area.

Objectives

• What are the current needs of surveillance professionals?
• What might technological solutions that incorporate machine learning look like?

Process

Before starting in on solutions, we needed to define our problem, our scope, and our end users. I facilitated design thinking exercises, including brainstorming and collaborative sketching, and encouraged the team to both keep an open mind to unusual ideas and empathize with the end user throughout the process.

Next, the team was reorganized into four subteams: AR/VR, 3D Map, Interface, and Code Limit Characterization.

AR/VR was charged both with creating a version of the surveillance solution that could be used in the future – when the technology is not as brittle nor uncomfortable to wear for an extended period, and making an immersive training solution for, e.g., disaster response.
3D Map used 3D modeling and camera positions to locate individuals in a video feed within a virtual environment. This is seen as closer to what is needed in the security profession to combat information overload, as software can easily be used to translate purely visual data into actionable information.
Interface needed to tie the project together in a way that was usable and intuitive to an end user. They used tools such as AxureRP and user interviews to develop a high-fidelity prototype of what might be put in front of the customer for this project. I helped these students use typical user experience research/design tools to figure out what is most helpful to the customer.
Code Limit Characterization (CLC) was organized around the idea that there are limits to facial recognition. The customer must understand those limits. In this project, the team spent a lot of time discussing the ethical considerations of using artificial intelligence for facial recognition and surveillance activity in different environments. The CLC team, specifically, looked at the limitations of the technology’s ability to detect faces in different environmental conditions. I helped the CLC design their experiment by specifying with them the variables they needed to control and measure. I also helped them run ANOVAs, effect size calculations, and multiple comparisons tests in PANDAS to analyze the data they collected. When preparing to present the results, I advised on the data visualization and interpretation pieces.

Conclusions

The overall project culminated in a capstone presentation that showcased the work of the student-interns to Boeing executives and SCDP family members. I had the opportunity to speak to my perceptions of the young professionals in the program and champion user-centered design to a company that is, excitingly, coming around on that idea in many ways. The project was a test of my leadership and mentoring ability, and it meant that I could teach some of the brightest up-and-coming engineers the value of user experience research and design.

Challenges

1. Communicating broadly
2. Leading a design cycle from start to finish
3. Serving as a mentor