Shreya Kashi

Working on ASAPP’s Generative Agent to improve customer interactions and overall user experience.

・​ Developing automated 2D-to-3D pipelines in Python for protein structure reconstruction from cryo-EM and volume-EM data, leveraging SVMs, CNNs, and tools like CryoGAN and CryoDRGN to advance drug discovery. ・​ Designing a signal quality module for SEM images to optimize data collection and contributing to research grant preparation. ・​ Developed a path optimization algorithm for efficient Cryo-EM microscope control.

・Worked on DARPA's Machine Common Sense initiative, with a focus on detecting passive violations of expectation in simulated video input using computer vision and machine learning techniques. Was among the top performing teams in the 2023 evaluation. ・ Used common deep learning frameworks like Pytorch for object detection, tracking and developed a comprehensive set of rules that model the basic laws of physics given a set of assumptions. For all tasks and datasets, created the data processing pipeline. ・Read research papers, presented their findings to my research group and translated them into technical projects.

・Designed, tested, and deployed three Angular/React-based web applications, integrating RESTful APIs while utilizing Agile methodologies. These solutions enabled customers to efficiently manage their energy subscriptions while significantly reducing administrative costs. ・ Implemented unit tests to maintain high test coverage; actively involved in defect triage to identify, track, and resolve issues. Utilized CI/CD pipelines in Azure DevOps for automated deployment and testing. ・Adhered to software engineering best practices, including modular object-oriented programming in JavaScript and Python, and version control using GIT. Participated in and conducted code reviews, providing feedback to improve code quality and foster team collaboration. ・ Documented technical processes and project milestones, creating detailed user guides and internal development resources. • Researched ways to improve customer experience through data analysis. • Participated in cross team functional planning meetings to analyze the requirements and provide suggestions for product development.

• Researched and implemented algorithms for contact-less breathing-rate detection to be used in the diagnosis of sleep-related disorders. Used Signal Processing techniques in Python to compare different algorithms for the same and illustrated the most accurate one. • Developed a signal quality module using linear regression which estimated the quality of the given signal with an accuracy of 94%. This module was integrated into the 'Sleepiz One' device, allowing medical professionals to discern signal reliability amidst noise.

Improved the accuracy of the previous model for water body detection in satellite images by 15% by employing super-resolution techniques to improve image resolution. Fine-tuned Mask RCNN model to detect water bodies in the processed images.

Taught computer and programming skills to underprivileged children. Managed a team of 15+ volunteers and planned the teaching program. Brainstormed and implemented ways to increase reach and participation from the children and the volunteers.

Extracted water bodies from satellite images using F-RCNN (Faster Region Convolutional Neural Network) and achieved an efficiency of 91%.

Worked on 'Malware detection using neural network' for which I used Cuckoo Sandbox to obtain the dataset.

Compared different kinds of underwater/pipeline surveillance robots based on the type of drive, depth of operation, type of camera and cost among other parameters.