Lakshmi Ramesh Email & Phone Number

Santa Clara, Ca, Us

United States

Indiana, United States

United States

• Devised a multimodal strategy with 4 types of MRI scans for 3D segmentation of 3 brain tumor classes• Utilized feature fusion to select 2 modalities that reduced training time by 60% and model size by 40%• Developed shifted-window cross-attention in PyTorch to attain 91% dice score for a transformer model• Designed a self-supervised ML pretraining strategy for 3D images using reconstruction and contrastive loss• Integrated the pretrained backbone in downstream segmentation to increase the dice score to 79%

Indianapolis, Indiana, United States

Assisting Qingxue Zhang in grading weekly assignments, projects and quizzes. Conducting weekly lectures and monitoring the class attendance for ECE 66200 - Pattern Recognition

Indianapolis, Indiana, United States

Assisted Professor Qingxue Zhang in grading weekly quizzes, projects and exams. Additionally conducted classes weekly and monitored class attendance for ECE 62900 - Neural Networks

Columbus, Indiana, United States

Cummins Inc., a global power leader, is a corporation of complementary business segments that design, manufacture, distribute and service a broad portfolio of power solutions.● Used NI controller with integrated FPGA to improve the design of ULD for ECU verification in LabView● Impact: Reduced human intervention by 70% and improved fault detection using the automated system● Evaluated the results of 20 simulations and documented a 4-month software proposal for a team of 8

Pune, Maharashtra, India

KPIT Technologies is a global partner to the automotive and mobility ecosystem for making software-defined vehicles a reality• Collaborated with the ADAS team to simulate and control various automotive sub-systems using AI• Employed YOLOv4 and Mosaic augmentation for multi-scale detection of pedestrians and crosswalks• Developed a MATLAB script that performs adaptive histogram equalization to boost the recall by 23%• Simulated automotive sub-systems using SIMULINK that leveraged CNN-guided control functions

Pune, Maharashtra, India

• Pioneered an NLP pipeline to perform emotion recognition in long paragraphs of customer reviews• Coproduced a Multi-task learning model that simultaneously classifies the arousal and valence classes• Conducted experiments using CNN, LSTM and transformer architectures in TensorFlow and PyTorch• Achieved an F-1 score of 77% by training an ensemble model composed of SVM, Naïve Bayes and BERT