Research systems used in real classrooms, web work for the VLED Lab, and earlier machine learning projects.
Active
Spandan
IIT Ropar · Jun 2025 – Present
A live classroom interaction platform that turns passive lectures into active learning. Spandan captures the teacher's speech in real time, transcribes it, generates relevant questions, and presents them to students through a dedicated portal for instant interaction and assessment.
Real-time speech capture and transcription
Automatic question generation from live lecture content
Student answer portal with live engagement tracking
Teacher dashboard with session analytics and participation monitoring
An adaptive classroom platform that makes teaching question-driven. After a brief topic introduction, students submit questions that shape where the class goes. Analysis of the collective questions gives teachers real-time insight into class mood, motivation and conceptual understanding; simple questions are answered automatically and complex ones reach the teacher's prioritised dashboard.
Analysis of student questions for mood, motivation and cognitive complexity
Smart triage: instant answers for simple questions, teacher escalation for complex ones
Real-time question flow with anonymous or identified submission
A multi-level peer assessment system with built-in quality control. Each student gets a unique ID per quiz for anonymity; teachers can download QR-coded IDs for offline use or let students upload through the portal. Statistical outlier analysis flags suspicious evaluations to Teaching Assistants, who can escalate to teachers.
Anonymous evaluation via unique student IDs, with optional QR codes
Statistical anomaly detection for suspicious peer evaluations
Escalation: peer → automated check → TA review → teacher
Rubric-based evaluation, analytics and complete audit trails
The public website of Vicharanashala, the Lab for Education Design at IIT Ropar, at vicharanashala.ai. I contribute features, content structure and maintenance, including the repository layout and documentation.
Data-driven pages editable without code
Events calendar, team and research pages
Clean repository structure and contributor documentation
A system that automatically recognises handwritten numerical digits on paper, detecting complete numbers so they can be processed for further analysis or use.
Digit detection on paper
Whole-number recognition for downstream use
PythonDeep LearningNeural Networks
Completed
Water Quality Analysis
JUIT · Jun 2022 – Jul 2022
Implemented several machine learning models and compared their accuracy to determine the best-performing model for water-quality classification.
Model comparison by accuracy
Water-quality classification
PythonMachine Learning
Completed
Lung X-ray Infection Segmentation
JUIT · Jan 2022 – Jun 2022
Generated segmentation masks of infections in the lungs from chest X-ray images using a deep learning model, aiding automated medical diagnosis.
