Passionate Computer Scientist with expertise in Intelligent Systems and Machine Learning, committed to lifelong learning and innovation. I continually seek new skills and experiences to create cutting-edge solutions that address real-world problems and drive technological advancements.
Concentration: Intelligent Systems & Machine Learning
Concentration: Intelligent Systems & Machine Learning
Thesis: "KIPPO - Koopman-Inspired Proximal Policy Optimization"
Minors: Cybersecurity, Business Administration
Applied Koopman theory to reinforcement learning for 6-60% performance gains over PPO with up to 91% variance reduction across continuous control benchmarks.
Modular federated learning framework supporting centralized, decentralized, and hierarchical topologies with plug-and-play algorithm integration scaling from edge to HPC.
Applied Koopman-inspired latent dynamics to surgical video, enabling multi-step prediction for robotic control and tool-tissue interaction analysis.
Self-supervised framework with scale augmentation and cross-scale consistency constraints eliminating need for aligned multi-scale satellite imagery; achieved state-of-the-art performance across remote sensing benchmarks.
Mentor: Sahil Tyagi
Architected modular federated learning framework supporting centralized, decentralized, and hierarchical topologies with plug-and-play integration of 11+ FL algorithms scaling from edge to HPC.
Designed and implemented abstraction layer and API contracts with standardized lifecycle hooks enabling seamless algorithm integration.
Integrated Ray and Hydra for distributed orchestration with mixed-protocol communication (MPI, gRPC), enabling topology and algorithm reconfiguration through YAML configuration.
Advisor: Dr. Hairong Qi, AICIP Lab
Developed KIPPO, applying Koopman theory to reinforcement learning for 6-60% performance gains over PPO with up to 91% variance reduction across continuous control benchmarks.
Co-developed Cross-Scale MAE with scale augmentation and cross-scale consistency constraints at structural and semantic levels, eliminating need for aligned multi-scale satellite imagery; achieved state-of-the-art performance across remote sensing benchmarks.
Applied Koopman-inspired latent dynamics to surgical video, enabling multi-step prediction for robotic control and tool-tissue interaction analysis.
Collaborated with start-up to build production diffusion-based design tool for interior, architecture, landscaping applications: FastAPI server, GPU inference on EC2, containerized deployment (ECS/ECR), S3 storage, CloudFront CDN.
Teaching assistant for COSC 420/527 Biologically-Inspired Computation with Dr. Catherine Schuman.
Developed Electron application (Quasar/Vue.js/TypeScript) for automated discovery and visualization of business logic relationships, integrating with Jira for security review workflows.
Built visualization tools for discovering and analyzing business logic relationships across systems.
Performed data processing, aggregations, and mappings by spidering through various sets of data queried using a REST API.
Assist with research and various ongoing projects, including the development of software and scripts to aid with the data collection, creation of visualizations, performing analysis of results, and writing and documenting research and findings.
Teaching assistant for Dr. Doowon Kim's Introduction to Cybersecurity course. Primary responsibilities include assisting students and answering questions they may have about the course materials, providing helpful feedback, grading quizzes and other assignments, exam proctoring, etc.
Assist with research on various ongoing projects, developing software and scripts to aid with collecting and analyzing data, creating visualizations, and interpreting results.
Topics: Security updates for JavaScript libraries, timeliness of security patches becoming available, the amounts and versions of vulnerable libraries still widely in use, crypto-jacking, certificate collection, phishing blacklists, and more.
Developed Android (AOSP) system services and framework components for enterprise fleet management system; integrated into production OS release enabling remote device configuration and monitoring.
Built Android client applications integrating devices with cloud-based fleet management platform, supporting remote configuration and OTA updates for tens of thousands of deployed touchscreen devices.
Developed customer-facing demos and proof-of-concept applications showcasing new platform capabilities for sales presentations and business development.