---

Work Experience

Coverbase Internship

Software Engineer Intern May 2025—August 2025

• Integrated Retrieval Augmented Generation (RAG) with Amazon Knowledge Base to create a vendor factsheet feature with modularized Tavily API + LLM queries, improving information retrieval speed by ~75%, deployed to all Coverbase customers
• Refactored multi-LLM Amazon Bedrock pipeline, resulting in faster execution and improved instruction-following by ~60% while reducing the API calls needed by ~50%

Technology Student Association

Software Developer December 2024—June 2025

• Top 12 of 100 teams at the National Technology Student Association Conference
• Created an end-to-end full-stack Flutter application with a Flask API backend to assist sustainable eating with real-time analysis through a custom locally hosted prompt-tuned Vision Language Model (LLaVA 13B)
• Developed a RAG pipeline with vector embeddings, reducing LLM hallucinations by ~90% and improving reliability

University of California, Santa Cruz Summer Internship Program (SIP)

Software Development Lead June 2024—August 2024

• Led software development for a 4-person team, building an end-to-end elderly fall detection system with Python, Android Studio, and Raspberry Pi for real-time video/audio streaming and processing.
• Enhanced fall detection accuracy by ~20% by integrating vision language models (Video-LLaVA, PaliGemma) with AlphaPose.
• Coordinated team efforts across hardware setup, ML integration, and Android app development; owned streaming pipeline + ML logic + notification system.

EduBeyond Internship

Web Developer Intern June 2023—Dec 2023

• Developed proficiency in web development using ReactJS, TailwindCSS, and MaterialUI
• Deployed 10+ features, such as password-protected classroom invites and a classroom settings page, to the Learning Management System

---

Research

Dual-Model Distillation for Efficient Action Classification with Hybrid Edge-Cloud Solution (NeurIPS Workshop on VLMs 2024)

• Created an optimized action-classification model through a hybrid edge-cloud solution by training a Multi-Layer Perceptron to route queries between various open-source Video-Language Models (ViLT, LLaVA)

Enhancing the Binary Classification of Wildfire Smoke Through Vision-Language Models (AIxSET IEEE 2024)

• Applied NLP innovations through Vision Language Models to camera imagery to enable early detection of wildfire smoke

Optimizing Large Language Models for Dynamic Constraints Through Human-in-the-Loop Discriminators (JMM 2025)

• Used Large Language Models, human-in-the-loop, and automatic prompt-tuning to create smart agents for travel planning