Trupti Sarje

Highly motivated Deep Learning Engineer with 3 years of experience in the computer vision domain. Proven ability to develop and implement effective solutions using deep learning frameworks. Adaptable and eager to learn new technologies, bringing a strong work ethic and collaborative spirit to any team.

Feasibility Study of Water Stress Detection in Plants using a High-Throughput Low-Cost System, 2020 IEEE SENSORS, Rotterdam, Netherlands, 2020

Netflix ChatBot - AWS Bedrock, Lambda, Llama, RAG, Langchain

Formulated activity recognition system to identify activities such as walking, climbing stairs for prosthetic limb. Used sequential timed signal from gyroscope and accelerometer as features for training the model. Utilized LSTM units from Keras framework for neural network. Achieved results for validation accuracy as 67%. 

Soybean Plant Water Stress Detector -  Scikit-Learn, Pandas, TensorFlow 

Devised water stress detector for Soybean plant using neural network, VGG16 as base model for transfer learning, to classify images into five levels of water stress. Resulting in the final accuracy of 80% over five classes. 

Segmentation based Plant Phenotyping - Detectron, Pytorch:

Formulated a deep learning model to identify of leaf, stem, and early leaf of plant using Detectron2 framework. Collected data, annotated it and designed data pipeline suitable for framework. Achieved result as mean average precision for segmentation 0.53 and for bounding box 0.57 at 0.5 IOU threshold.