Noteboost
Project Info:
Team Size: 1 Director of Research, 1 CS Professor, 1 Project Manager, 2 Engineers
My roles and responsibilities:
- Building the frontend of the student view (React)
- Implementing backend, including authentication (Node.js + Passport.js)
- Implementing version 2 (now deprecated) of the NLP algorithm for automated short-answer grading
- Building data pipeline and preprocessing
Technologies/Tools used: React Node.js MongoDB Javascript Tailwind CSS Python
Project Timeline: January 2021 - Present
Noteboost is an automated short answer grading system intended to be used to grade chartnotes for medical school students. The system is currently deployed in pilot stage for a top medical school in the US and has accumulated around 60 users to date.
The primary objective of the system is to provide real-time feedback to medical school students on the accuracy of their medical chartnotes in simulation study.
The work that was produced as a part of this project was presented at the International Meeting for
Simulation in Healthcare in Los Angeles, Jan 2022.
We have also had the opportunity to collect feedback from 5 user surveys, as a chance to continually improve the UI/UX of the system as well as iterate on the performance of the NLP module.
The student view consists of feedback and automated grading on four different kinds of checklists (History, Physical Exam, Differential Diagnosis and Tests), and a screenshot of one of these views has been included below:
This view displays the student answer, the checklists that the NLP algorithm detected in the student answer and the ones that it did not. The UI for this view was developed as a result of several UI iterations and user feedback. #Include info here
We are currently working on a paper that outlines the details of the NLP algorithms, which will be shared later this year.
Project Outcomes:
The biggest outcome from this project was the fact that we were able to deploy the system in a real world setting and provide value to the users. It was also really instructive to obtain and learn from user feedback that was provided by our users. As engineers and product designers, it is common to encounter blind spots from the user's perspective, but the ability to incorporate feedback, correct mistakes and improve the user experience of a software product was a really instructive experience. The real reward for our team was that at the end of the product iteration lifecycle, we found that majority of our users loved using the system and it was providing them with real value.
Learnings:
An important learning from this project was the experience of building a production ready system in deployment. The experience of designing a web-based end-to-end ML powered system was very instructive from a full-stack web development, as well as a ML system design perspective.