Rushil Goomer Email & Phone Number

Designed and implemented a graph-based evaluation framework for AI agents to track functional and non-functional milestones during conversations, ensuring adherence to business workflows through dynamic milestone dependency graphs.Introduced new evaluation metrics for AI agents by calculating progress rates, goal completion, and workflow compliance, bridging the gap between traditional QA evaluations and business-oriented performance benchmarks.

Canada

Ensure Scalable Architecture: Lead the design of a scalable and secure backend infrastructure capable of handling high user traffic and supporting future platform growth.Integrate Core Features: Supervise the implementation of essential social media functionalities such as user profiles, content sharing, messaging systems, and real-time notifications to deliver an engaging user experience.

Winnipeg, Manitoba, Canada

Funded by: Natural Sciences and Engineering Research Council of Canada· Developed Machine Learning models on AWS SageMaker, with a pixel-to-pixel regression approach akin to SRCNN, tailored for enhancing weather maps for North America.· Employed AWS S3 for secure data storage and Scikit-learn Pipelines to streamline preprocessing steps, ensuring seamless data transformations and reproducibility.· Successfully trained ML models including, XGBoost, Convolutional Neural Networks (CNN),… Show more Funded by: Natural Sciences and Engineering Research Council of Canada· Developed Machine Learning models on AWS SageMaker, with a pixel-to-pixel regression approach akin to SRCNN, tailored for enhancing weather maps for North America.· Employed AWS S3 for secure data storage and Scikit-learn Pipelines to streamline preprocessing steps, ensuring seamless data transformations and reproducibility.· Successfully trained ML models including, XGBoost, Convolutional Neural Networks (CNN), and Graph Neural Networks (GNN) with custom loss functions, and seamlessly deployed them using SageMaker Endpoints for operational efficiency.· Utilized SageMaker’s built in Experiments, MLOps tools and Model Registry to track hyperparameter tuning and generate insights about the training process.· Achieved significant improvements with a 17% reduction in Absolute Errors and 47% reduction in Median Error. Show less

India

- Developed an image processing application to detect electrical components on a custom PCB’s using YOLOv5, aiming to enhance manufacturing efficiency.- Curated and annotated a dataset of 1500 images using CVAT and stored on AWS S3 bucket to train the model effectively.- Trained a YOLOv5 model on AWS Sagemaker, optimizing hyperparameters and leveraging transfer - Created a Streamlit application to visualize and test YOLOv5 model predictions and implemented Docker containers to… Show more - Developed an image processing application to detect electrical components on a custom PCB’s using YOLOv5, aiming to enhance manufacturing efficiency.- Curated and annotated a dataset of 1500 images using CVAT and stored on AWS S3 bucket to train the model effectively.- Trained a YOLOv5 model on AWS Sagemaker, optimizing hyperparameters and leveraging transfer - Created a Streamlit application to visualize and test YOLOv5 model predictions and implemented Docker containers to encapsulate the application environment.- Achieved out standing performance metrics with a precision of 0.85 and recall of 0.86, demonstrating robust detection capabilities. Show less

India

Worked on detecting fluoride concentrations in water samples using UV Spectroscopy with data from carbon dot-based sensors.Collected data and extracted features from the fluorescence spectra, including peak intensities and emission profiles.Developed supervised machine learning models like regression and boosted trees to accurately predict fluoride concentrations based on fluorescence emission data.Improved the accuracy and robustness of the detection process, achieved… Show more Worked on detecting fluoride concentrations in water samples using UV Spectroscopy with data from carbon dot-based sensors.Collected data and extracted features from the fluorescence spectra, including peak intensities and emission profiles.Developed supervised machine learning models like regression and boosted trees to accurately predict fluoride concentrations based on fluorescence emission data.Improved the accuracy and robustness of the detection process, achieved MSE of $0.6 mg/L^2, contributing to reliable and efficient fluoride monitoring. Show less

· Developed a machine learning-based non-invasive blood pressure measuring device.· Collected data for 137 patients and persisted ECG and PPG sensory data on to AWS I.O.T. services.· Implemented wavelet transformation and regression analysis for blood pressure estimation.· Brought down the mean error to +/−3.854 mmHg which is acceptable as per the British Hypertension Society.