A curated collection of research projects, experiments, and engineering work spanning AI, ML, and computer vision.
Document Scanner Web App with custom Named Entity Recognizer (NER) for business card extraction. Built end-to-end pipeline: data labeling (BIO tagging), text parsing, Displacy rendering, and bounding box annotations.
View DetailsRanked 9th of 400+ teams at IndoML 2024 (BITS Pilani Goa). Multi-task classification using BERT, LLaMa-2-7B, and GPT-J. Integrated FAISS-based data augmentation with adaptive learning for real-world e-commerce.
View DetailsMajor Project. Implemented Bi-ANN, Bi-LSTM, and BERT for sequential classification of medical abstracts. Achieved 92% accuracy via fine-tuning and optimization. Full preprocessing & feature extraction pipeline.
View DetailsBuilt at AMURoboclub. Designed a 3D robotic arm to sort objects by color. Used OpenCV for robust hue-based classification and Raspberry Pi for precise mechanical control of arm movements.
View DetailsIntuitive web app to predict housing prices. Trained a regression model on real estate features, deployed via pickle for fast inference. Designed a clean front-end using HTML/CSS for smooth UX.
View DetailsDeep learning model with transfer learning for food image classification. Enables accurate recognition of diverse foods — applicable for diet tracking, calorie estimation, and nutrition analysis.
View DetailsTwo-agent CNN-Transformer for indoor REM. Combines local feature extraction (CNN) with global context (Transformer). Auxiliary LoS and antenna pattern inputs. 15%+ improvement over SOTA in MSE.
View PublicationPublished at IEEE MASS 2024 (Seoul). Real-time vehicle detection with YOLOv8s for adaptive traffic light control. Achieves mAP@0.5 of 85.8% across 5 vehicle categories.
View Publication