Mansi Singh

Architected a production-grade multi-agent AI research lab orchestrating 8 specialized AI agents to conduct, verify, and synthesize academic research on any topic. Implemented agentic workflows with human-in-the-loop checkpoints for research planning, literature review, fact-checking, and report generation.

Built an end-to-end computer vision pipeline combining YOLOv8 object detection and ResNet50 classification on the IP102 dataset (45,095+ images across 102 pest classes). Engineered imbalance-aware training using data augmentation, transfer learning, and fine-tuning, achieving 68.58% classification accuracy and 77.6% mAP@0.5; integrated TensorFlow Lite models and LLM-assisted decision support into

Developed an interactive OCR correction platform for scanned documents using confidence-aware generative AI with human-in-the-loop error correction workflows. Implemented image preprocessing and real-time OCR review/correction via a desktop interface, achieving 75% word-level accuracy and 73% character-level accuracy on noisy datasets.

Grade: CGPA: 7.2

Pursuing a B.Tech in Computer Science & Engineering with a minor specialization in Machine Learning. Working on edge and cloud AI/ML solutions including computer vision, NLP, generative AI, and multi-agent systems (CGPA: 7.2).

Grade: 83%

Completed Intermediate at Sant Atulanand Convent School, achieving 83%.

Grade: 90%

Completed Matriculation at Happy Model School, achieving 90%.