• Conducting feasibility analysis and pre-seed due diligence to validate POC designs, culminating in the development of working demo of the product.• Developing a Knowledge Graph using Neo4j and a vector store with Weaviate to semantically organize information, and engineering a RAG-based architecture with container-deployed large language models (LLMs), orchestrated data flows using Kafka, and custom API endpoints.

The goal of this project is to utilize cutting-edge generative AI, machine learning, statistical modeling, and simulation techniques to effectively identify ransomware attacks. Employing generative models involves simulating a time-series-based snapshot of a ransomware attack and generating the following artifacts:1. File-based datasets: Utilizing GenAI models to enhance existing file-based datasets such as GovDocs1 [3] and NapierOne [4] by creating a new dataset.2. User-activity datasets: Utilizing datasets like CLUE-LDS [5] to establish a file-activity pattern for training machine learning models.• Created synthetic datasets for ransomware analysis using simulation techniques and generative AI models (GPT 3.5, Replicate.ai, Easy Diffusion).• Developed a time-series analysis model with Explainable AI to detect abnormal user activities, achieving a 95% F1-score and actively in use within the company.• Recognized with the ”Best in Show” award among 16 exceptional projects.

• Conducted research in collaboration with the US Forest Service and University of Washington to estimate visitation in high-risk wildfire-susceptible landscapes that span 208,000 acres.• Findings through geospatial analysis and visitation models contributed to informed policy decision-making regarding accelerating fire treatments in high-priority locations.

• Streamlined applicant tracking system requirements through stakeholder feedback consolidation, ensuring alignment of technical solutions with business requirements.• Leveraged PowerBI to read, manipulate, conduct EDA and model data, creating insightful visualizations for strategic KPI-driven decision-making.

• Integrated a Spring Boot Microservice with a CI/CD pipeline and CCM to decrease development time by 67%.• Designed Splunk dashboards to visualize communication metrics covering 25+ events and a total of 400,000 emails, text messages and notifications sent per day.

• Deployed a Spring Boot service to communicate order status to over 3 million customers, reducing customer service calls in Walmart Canada by over 50%.

• Leveraged machine learning techniques for object detection and localization in a drone dataset.• Led a comparitive analysis between the RetinaNet and Faster RCNN to evaluate that although Faster RCNN was 100% faster than RetinaNet, it was 50% less precise in its detection.

• Assisted research in designing a CNN model that performs both Vehicle Classification and Vehicle Re-identification in Python with the use of OpenCV and PyTorch