vCard - Personal Portfolio

About me

I am an enthusiast of explainable AI and medical imaging, with extensive experience in computational modeling, multimodal ML, and AI platforms, driven by a passion to create transparent, impactful technology at the intersection of healthcare, research, and global sustainability.

My expertise spans Python, R, MATLAB, and C++, with a strong foundation in machine learning frameworks like TensorFlow and PyTorch. I have a proven track record of leading cross-functional teams to deliver innovative solutions in healthcare and environmental sustainability.

What I'm Building

Explainable AI Systems

Designing interpretable and transparent AI models that bring trust to healthcare and real-world decision-making.
Machine Learning Research

Exploring novel architectures and cognitive-inspired algorithms for AI reasoning and memory simulation.
AI-Driven Applications

Building intelligent apps that blend computer vision, NLP, and interactive explanations for medical imaging and beyond.
Innovation & Product Design

Merging deep tech with creativity—transforming bold AI concepts into products that solve real human problems.

Resume

Education

VANDERBILT UNIVERSITY
2023 — 2027
James C. Seuss Scholarship, Dean’s List(each semester), Nichols Humanitarian Fund 2025, Commodore Cup Finalist, Vanderbilt Blockchain
Hong Kong University of Science and Technology
2025 — 2026
Exchange Student, Department of Computer Science and Engineering
Courses: Deep Learning for Computer Vision, Machine Learning, Statistics for Machine Learning, Artificial Intelligence
Kathmandu Model College (KMC)
2019 — 2021
Merit Scholar(Only top 1%), National Astronomy Olympiad, Top 25 in National Math Olympiad

Experience

Summer Intern, UBICOMP Lab
May 2025 — Jul 2025
• Selected from 4,000+ applicants as one of 40 interns; ran user studies (12 participants) comparing decision tree vs. logistic regression explanations. Found most relied on linear arithmetic shortcuts while struggling to recall deeper trees (≥3 nodes).
• Built a case-based retrieval system that reused prior examples for forward and counterfactual reasoning, raising behavior-prediction accuracy by 35%.
• Designed a framework that stored simple heuristics (e.g., “higher feature 4 → Type 2”) instead of full weights, reflecting how users compressed explanations. Showed DTs were faster but harder to memorize than LRs.
Undergraduate Research Assistant, Vanderbilt University
Dec 2023 – Dec 2024
• Automated multimodal data capture (video, gestures, speech) using OpenCV + librosa, reducing manual annotation time by 40% and enabling faster iteration for behavioral analysis.
• Designed and deployed SQL-backed dashboards (Python, R, PostgreSQL) that provided real-time student engagement analytics, helping instructors tailor interventions dynamically.
• Containerized data pipelines into Dockerized Flask microservices, ensuring reproducibility and scalable deployment across different lab environments.
Summer Research Assistant, Vanderbilt University
May 2024 - July 2024
• Trained and optimized TensorFlow/Keras neural networks to model cavitation bubble dynamics, achieving 98% accuracy on 20k+ experimental data points through advanced hyperparameter tuning and data augmentation.
• Integrated Runge–Kutta numerical solvers into ML workflows, reducing simulation runtime by 30% while maintaining physical accuracy.
• Developed a hybrid experimental + computational dataset calibration pipeline, which improved model robustness and reproducibility across trials.
• Validated predictions against experimental benchmarks, demonstrating alignment between theoretical models and physical observations, and paving the way for more accurate fluid dynamics simulations.