Projects & Systems

Real-time face recognition attendance system with automated identity recognition, CSV-based attendance tracking, and Flask administration.

• Built real-time face detection using OpenCV Haar Cascade.
• Implemented KNN-based face recognition using registered face images.
• Automated attendance capture with name, roll number, and timestamp.
• Persisted daily attendance records using Pandas and CSV files.
• Developed a Flask interface for registration, attendance viewing, and model management.
• Automated KNN retraining whenever new users are enrolled.

An online judge and automated test-case evaluation system designed to safely compile and execute untrusted code in sandboxed environments.

• Designed isolated and sandboxed execution environments for compiling and running untrusted code.
• Built backend REST APIs and an asynchronous judging pipeline for high throughput.
• Implemented automated test-case evaluation for functional correctness, boundary conditions, and edge cases.
• Enforces strict runtime, memory limits, and process isolation.
• Added structured logging for debugging failed submissions and performance telemetry.
• Managed and deployed on a self-managed Linux VPS with Docker containerization.

A diffusion-based generative computer vision pipeline for image generation, latent space conditioning, and processing.

• Diffusion-based image generation and latent space conditioning.
• Fine-tuning pipelines with LoRA and custom vision datasets.
• PyTorch model training, evaluation, and latency benchmarking.
• Integrated with Hugging Face Hub for weights distribution.

A machine-learning pipeline for analyzing academic data and predicting student performance outcomes.

• Comprehensive exploratory data analysis (EDA) and feature engineering.
• Regression and classification modeling with scikit-learn algorithms.
• Evaluated performance with cross-validation, RMSE, and F1 metrics.
• Structured automated pipelines for clean data transformations.