Gradient.ai: Student Dashboard
Gradient.ai: Graded Assignment Page & Overall Feedback, Student View
Gradient.ai: Assignments Page, Student View
Gradient.ai: Per Question Analysis and Feedback, Student View
Gradient.ai: Landing Page

About the Project: Gradient.AI

Inspiration:
The inspiration for Gradient.ai came from observing the challenges educators face in grading large volumes of student answer scripts, which often leads to delays or rushed grading to satisfy the deadlines. More often than not, manual grading is time-consuming, bias-prone, and inconsistent. With the power of LLMs and generative AI, we have the capability to understand and enhance, make grading faster, fairer, and insightful.

What We Learned:

How to apply Natural Language Processing (NLP) techniques to analyze text-based answers. Use of Gemini APIs,
How to design a rubric-based scoring system that the AI can interpret.
Integration of AI models with web applications for real-time feedback.
The importance of balancing accuracy with efficiency in automated evaluation systems.
Full Stack ML Solution built on Python and NextJS

How We Built It:

Collected sample student answer scripts and corresponding rubrics as generated through the semester.
Used the OCR to transform the documents into text. Input format is flexible across LaTex, Markdown, and Handwritten text documents.
Tune the Gemini Model to evaluate answers against rubric criteria: correctness, completeness, and clarity. Utilisation of different personas while grading, use of a knowledge base for feedback and generating insights through context.
Built a scoring engine to assign per-question and total scores, along with actionable feedback.
Integrated the AI engine into a user-friendly web interface for educators to upload answers and rubrics easily.

Challenges Faced:

Handling ambiguous or creative answers that do not strictly follow the rubric.
Ensuring the AI assigns consistent scores across different students and questions.
Optimizing the system to process large batches of scripts efficiently.
Designing feedback that is both accurate and helpful without overwhelming the student.

Conclusion:
Gradient.ai transforms traditional grading into an automated, efficient and consistent process, helping educators focus more on teaching and less on administrative tasks.