Resume

July 2022 - Present

Collaborated with Physicists from NCNR[NIST], PML[NIST] and UMD on building workflows for data from novel INFER method.
Demonstrated the effectiveness of synthetic data obtained from Data-driven simulations for training AI based segmentation of Neutron Images using a checkerboard-based class-balancing strategy. Added synthetic scene generation and augmentations for anticipated future experiments. This addressed challenges caused by data scarcity due to prolonged neutron source shutdowns.
Implemented validated webforms to standardize generation of complex experimental metadata files and for ai-model-cards.
Investigated the impact of mixing physics based and data-driven synthetic data on model training and inference.
Containerized and deployed plugins and workflow to the Web Image Processing Pipeline (WIPP) developed at NIST.

December 2019 - July 2022

Conducted 3D image segmentation on confocal stacks of genetically modified human Retinal Pigment Epithelium cells.
Formulated objectives, workflows and algorithms with biological experts (NEI/NIH) and computer scientists (NIST).
Extracted Individual cell objects and selectively stained organelles to obtain biologically relevant shape and organization metrics.
Analysis confirmed and expanded recent conclusions of biological effect of inhibitor vs enhancer of ciliogenesis in RPE.

January 2017 - December 2019

Applied clustering (e.g. k-NN, GMM) and dimensionality reduction (e.g. PCA, CCA) with a focus on interpretability of skeletal joint coordinate data for summarizing trajectories of exercise activity for classification/recognition, diagnosis, and guidance.
Implemented client-server based synchronized multi-Kinect™ system that can improve tracking of landmark joints with body segment occlusion and sensor jitter. Live classification demo can classify images based on preceding 2-4 seconds of data.
Demonstrated classification using Neural Networks, Support Vector Machines, Permutation Entropy and Mutual Information.
Forecasting using ARIMA and Regression methods (SVR, Harmonic Regression) and HMMs. Work presented in 6 poster sessions.

September 2014 - December 2016

Investigated factors influencing printability including coffee ring effect, contact angle hysteresis, porosity, absorption, etc.
Prepared vinyl formulations, cast films, heat treated and laminated, printed using large format printers before testing.
Investigated and characterized parameters such as end-groups and chain length additive small molecules structures, or topcoats to observe surface migration and blooming of additives. Improved additive choice for further testing at pilot scale.
Other areas of exploration included polarized FTIR for machined polymer films, effect of evaporation on droplet attraction, etc.

Welcome