Auto-Teach

This is the problem screen, you can put your approaches into the text box.
This is an example having a wrong answer but correct approach
This is an example having a correct answer

Inspiration

The inspiration for our Auto-Teach project stemmed from the growing need to empower both educators and learners with a self-directed and adaptive learning environment. We were inspired by the potential to merge technology with education to create a platform that fosters personalized learning experiences, allowing students to actively engage with the material while offering educators tools to efficiently evaluate and guide individual progress.

What it does

Auto-Teach is an innovative platform that facilitates self-directed learning. It allows instructors to create problem sets and grading criteria while enabling students to articulate their problem-solving methods and responses through text input or file uploads (future feature). The software leverages AI models to assesses student responses, offering constructive feedback, pinpointing inaccuracies, and identifying areas for improvement. It features automated grading capabilities that can evaluate a wide range of responses, from simple numerical answers to comprehensive essays, with precision.

How we built it

Our deliverable for Auto-Teach is a full-stack web app. Our front-end uses ReactJS as our framework and manages data using convex. Moreover, it leverages editor components from TinyMCE to provide student with better experience to edit their inputs. We also created back-end APIs using "FastAPI" and "Together.ai APIs" in our way building the AI evaluation feature.

Challenges we ran into

We were having troubles with incorporating Vectara's REST API and MindsDB into our project because we were not very familiar with the structure and implementation. We were able to figure out how to use it eventually but struggled with the time constraint. We also faced the challenge of generating the most effective prompt for chatbox so that it generates the best response for student submissions.

Accomplishments that we're proud of

Despite the challenges, we're proud to have successfully developed a functional prototype of Auto-Teach. Achieving an effective system for automated assessment, providing personalized feedback, and ensuring a user-friendly interface were significant accomplishments. Another thing we are proud of is that we effectively incorporates many technologies like convex, tinyMCE etc into our project at the end.

What we learned

We learned about how to work with backend APIs and also how to generate effective prompts for chatbox. We also got introduced to AI-incorporated databases such as MindsDB and was fascinated about what it can accomplish (such as generating predictions based on data present on a streaming basis and getting regular updates on information passed into the database).

What's next for Auto-Teach

Divide the program into two mode: instructor mode and student mode
Convert Handwritten Answers into Text (OCR API)
Incorporate OpenAI tools along with Together.ai when generating feedback
Build a database storing all relevant information about each student (ex. grade, weakness, strength) and enabling automated AI workflow powered by MindsDB
Complete analysis of student's performance on different type of questions, allows teachers to learn about student's weakness.
Fine-tuned grading model using tools from Together.ai to calibrate the model to better provide feedback.
Notify students instantly about their performance (could set up notifications using MindsDB and get notified every day about any poor performance)
Upgrade security to protect against any illegal accesses