Work Experience

Tech: Python, PyTorch, CUDA, multi-GPU optimization, multimodal datasets, AI agent frameworks

• Developing AI agents for advanced multimodal research, integrating image, text, and video understanding tasks.
• Designed and implemented chunking algorithms to efficiently preprocess and manage large multimodal datasets.
• Researching and prototyping optimized deployment strategies for sequential and parallel AI agents across multi-GPU systems, improving scalability and throughput.

Tech: C++, Python, Data Structures & Algorithms, Teaching & Mentorship

• Assisting in teaching Algorithms course to undergraduates, covering fundamental and advanced topics in design and analysis.
• Conducting recitation sessions and problem-solving workshops to reinforce theoretical concepts with practical coding exercises.
• Supporting students through office hours, one-on-one mentoring, and providing feedback on assignments and projects and grading.
• Collaborated with faculty to create teaching materials and practice problems, improving student exam preparedness.

Tech: Workday, Python, ELK Stack (Elasticsearch, Logstash, Kibana), AWS, JavaScript/HTML/CSS

• Built automation scripts for Workday client processes, streamlining student enrollment and salary calculations and reducing administrative effort.
• Implemented structured logging with the ELK Stack, introducing correlation IDs and custom middleware for request tracking, improving end-to-end traceability and cutting issue resolution time by 60%.
• Developed and maintained frontend components for SCU websites, enhancing UI/UX design and creating new pages for academic and administrative use.

Tech: Python, CLIP, Mediapipe, CUDA, AWS (Lambda, SQS, API Gateway), GPU optimization

• Designed and implemented a GPU-optimized video processing pipeline using Python, CLIP, and Mediapipe for a 300GB+ dataset.
• Parallelized frame-level operations, reducing vector embedding generation time from 50+ hours to ~21 hours.
• Deployed trained models on AWS cloud, integrating SQS queues, Lambda, and API Gateway for scalable query handling.
• Delivered an end-to-end system optimized for Meta VR headset integration, improving latency and reliability of video-based research tasks.

Tech: Java, Swing UI, server programming, OOP, Agile

• Taught core Data Structures and Algorithms, from arrays and linked lists to advanced graphs and heaps, using Java.
• Guided students in Object-Oriented Design Principles and project development with Swing UI and server-based programming.
• Supervised Low-Level Design (LLD) projects, emphasizing clean code and real-world applicability.
• Trained students on Software Engineering principles and Agile methodologies, bridging theory with practice.

Tech: PyTorch, TensorFlow, Python, multi-GPU training, YOLO models

• Explored advanced PyTorch and TensorFlow methods for image processing and performance optimization across varied compute setups.
• Developed automation scripts to manage multi-GPU training pipelines, enabling efficient large-scale dataset experiments.
• Optimized YOLO models for hand-sign detection, producing best-performing weights and benchmarking accuracy improvements.
• Co-authored a paper in the SSRN Journal with 700+ downloads and multiple citations, highlighting reproducible results.
• Open-sourced all model weights and source code, contributing a valuable resource to the research community.

Tech: Python, scikit-learn, NumPy/Pandas, signal processing, XML automation, deployment

• Researched and trained ML models for 3-lead ECG rhythm prediction, achieving ~93% accuracy with medically graded sensor integration.
• Contributed to a novel 3-lead ECG dataset, including collection standards and labeling guidelines.
• Built XML automation pipelines for data extraction and preprocessing to accelerate experimentation.

Tech: C++, custom libraries, IoT, system performance optimization

• Developed C++ libraries for system software, optimizing hardware–software interaction.
• Reduced IoT transfer latency from 5–7s to milliseconds, drastically improving end-to-end system performance.