The 3-Step Workflow: Upload → Evaluate → Publish
The DASES workflow is designed to minimise faculty time spent on repetitive tasks while keeping them in full control of the final evaluation. The system operates on a straightforward three-step model: Upload (where exam materials and student responses are ingested), Evaluate (where the AI pipeline processes the handwriting and scores against rubrics), and Publish (where faculty review the AI's work and release results to students). This workflow replaces the traditional, time-consuming cycle of manually reading, deciding scores, calculating totals, and manually writing feedback for every individual student.
Step 1: Paper Setup and Rubric Generation
Before answer sheets can be graded, DASES needs to understand the exam. Faculty begin by creating a paper in the system. They can upload an existing question paper PDF, and DASES will extract the questions automatically. Next, faculty provide a model answer for each descriptive question. DASES uses this model answer to automatically generate a detailed grading rubric — breaking the expected response into specific criteria (e.g., "Correct formula application," "Accurate final calculation"), assigning weights, and defining partial credit rules. Faculty review this AI-generated rubric, making any necessary adjustments to weights or adding alternative acceptable approaches. This entire setup process typically takes under 10 minutes.
Step 2: Answer Sheet Upload and AI Processing
Once the exam has been administered, the physical answer booklets are scanned into PDF format. Faculty upload these PDFs to DASES in bulk. If a PDF contains multiple students' papers scanned sequentially, DASES automatically identifies the boundaries and splits the document into individual student submissions. Once uploaded, the core AI pipeline takes over. It segments the pages to find specific answers, reads the student's handwriting using advanced recognition models, maps the text to the appropriate question, and evaluates the semantic meaning of the response against the established rubric. DASES processes up to 500 sheets in parallel, evaluating the entire batch in approximately 25 minutes.
Step 3: Faculty Review and Report Generation
The AI does not publish results autonomously. After processing is complete, faculty are presented with a review dashboard. This dashboard shows the AI-assigned score for every question across all students. DASES flags answers where the AI had lower confidence (for instance, if the handwriting was extremely illegible or the answer diverged significantly from the rubric but still contained relevant keywords). Faculty can spot-check these flagged answers, view the student's original handwriting alongside the AI's evaluation, and override the score or feedback with a single click if their professional judgment differs from the AI. Once satisfied, faculty click "Publish" to finalize the results.
What Faculty See: The Teacher Dashboard
The teacher dashboard provides a comprehensive command center for exam evaluation. It offers a macro view of class performance, including score distributions, average scores per question, and common areas where students lost marks. It also offers a micro view, allowing faculty to drill down into a specific student's paper to see exactly how the rubric was applied to their handwritten response. The dashboard highlights questions that proved exceptionally difficult for the class, enabling faculty to adjust their teaching strategy or revise the rubric if a question was overly ambiguous.
What Students See: The Student Portal
Upon publication, students receive a notification and can log in to the DASES student portal. Instead of just seeing a final grade, they see a highly detailed breakdown. For every question, they can view the original image of their handwritten answer, the score they received, the maximum possible score, and — crucially — the specific rubric criteria applied. DASES generates written feedback explaining exactly why marks were awarded or deducted based on the rubric. If a student wrote an incomplete derivation, the feedback will explicitly state which step was missing. Students also have the option to download a branded PDF version of this comprehensive report.
Behind the Scenes: The AI Pipeline
The speed and accuracy of DASES rely on a complex, multi-stage AI pipeline operating invisibly in the background. It combines state-of-the-art computer vision (for page layout analysis and handwriting recognition) with large language models fine-tuned for educational assessment (for semantic evaluation and feedback generation). The system is specifically calibrated for the nuances of student exam papers: it can handle crossed-out text, margin notes, arrows indicating continued answers, and mixed cursive-print writing styles. The evaluation engine doesn't just look for exact keyword matches; it understands synonyms, paraphrasing, and conceptually equivalent answers.