SmartOH

Student rooms
Student question
Student queue
TA rooms
TA queue and student selector
our workflow

Inspiration

As students and TAs ourselves, we've experienced the frictions caused by the current way we conduct office hours. The most significant problems that we identified include:

Current solutions for office hours queues are opaque and unintuitive
The largest classes (e.g. entry-level CS courses) also tend to be those with the most students in need of help, and the lowest variation of questions asked during office hours
Each course uses its own office hours tool, with no reminder system or centralized calendar.

With these problems in mind, we set out to improve students' experience and thoughtfully optimize the time of TA interaction.

What it does

SmartOH provides a comprehensive interface to allow TAs to coordinate office hours and optimize their time. The unique features of our platform include:

TAs have the ability to move people into their rooms based upon the similarity of students' questions (built using BERT).
Students have an unparalleled level of visibility into the length of the queue and the expected wait time as the platform records the length of time spent for each question by TA.
Students have a single point of access for all of their office hours, so are more likely to be aware of and attend office hours.
Professors have an effective tool to measure the speed and quality of their TAs' office hours, and have access to a valuable pool of information that could be used to drive improvements.
Transcripts of each conversation between a TA and a student are summarized and converted into LaTeX, and may be shared with the student or the entire class if the question covered is useful.

How we built it

The backend was built using Node, and the frontend was built with React and CSS. We used web sockets for the live components of the site.

We used transcription APIs from Google, and we used BERT and PyTorch for similarity analysis. We used GPT 3.5 Turbo for some summarization and theme identification. We deployed Python microservices via FastAPI interface.

Challenges we ran into

We spent a lot of time working on the backend to make sure that the live components worked as anticipated. We also spent a great deal of time making sure that the UX was strong, as that is an important component of a platform like this.

Accomplishments that we're proud of

Our team has a very diverse group of skillsets, and we are proud of our ability to combine all of our competencies to create a platform that we are proud of.

What we learned

We learned how to work with websockets in constructing a live application such as SmartOH.

What's next for SmartOH

Going forward, we would love to use the Zoom SDK to integrate online meetings into our platform. We looked into it for the hackathon, but it looked like it was oriented towards commercial use so was infeasible for us to implement. We believe that this would be an extremely compelling offering if we integrated a communication service with it, and we will be continuing to look into this.