Rina Kim

Selected Works

00

BMW Adaptive Generative UI

2025 June - 2026 January · BMW Group Technology Office · UX Engineer Intern

Overview

• Developed an adaptive generative UI system for BMW vehicles that dynamically adjusts interface elements based on driver behavior, environmental conditions, and personal preferences.

Key Contributions

• Developed a robust React-based orchestration layer to synchronize real-time telemetry between asynchronous sub-agents and the end-user interface, ensuring seamless human-machine interaction.
• Engineered a modular multi-agent architecture that delegates intensive rendering tasks to specialized sub-agents, eliminating synchronous execution stalls and ensuring high-fidelity, fluid interface responsiveness.

Topics Researched

• Generative AI
• System Design
• LLM

• Layered Architecture Diagram

• GenUI architecture illustrates a tiered synthesis pipeline that orchestrates contextual processing to deliver a real-time adoptive interface.

• Synthesis pipeline that transitions from raw contextual processing to a fluid, high-fidelity adoptive interface.

01

SmaSH Lab | Proactive Agent

2024 September - 2025 August · Carnegie Mellon University SmaSH Lab · Research Assistant

Overview

• Researched & built a semantic classification pipeline for adaptive multimodal systems that isolates relevant user intent from environmental noise and linguistic variations, achieving >90% response accuracy.

Key Contributions

• NLP pipeline development
• User intent prediction model
• Real-time response generation

Impact

• Through adaptive multimodal systems, this project was designed to isolate user intent despite linguistic variations or background interference both common failure points for standard voice assistants. The semantic classification pipeline effectively filters 'human signal' from 'environmental noise.'

02

AI Trend Forecasting Tool

2024 January - 2024 August · Carnegie Mellon University | Surefront · UX Researcher & Developer

Overview

• Built B2B tool designed to bridge the gap between high-level data analytics and creative execution. It leverages real-time data and AI to empower fashion professionals from designers to merchandisers to make informed, proactive decisions.

Key Features

• Dynamic Data Integration: Built to process and visualize real-time growth metrics, search volumes, and month-over-month performance curves.
• Agentic AI Implementation: The Personal Fashion Assistant democratizes data. Instead of digging through three layers of menus to find "shoes in France," a user can simply type: "What footwear is trending in Paris right now?"
• Advanced Analytics & Visualization: Built to process and visualize real-time growth metrics, search volumes, and month-over-month performance curves.

Impact

• 75% Reduction in Research Latency: By consolidating real-time Google Trends APIs and social media scraping into a single HMI, the time required to validate a trend aesthetic against hard search data was reduced from 4 hours of manual cross-referencing to less than 60 seconds of automated dashboard interaction.
• 90% Decrease in Interaction Cost: The implementation of the "Agentic AI Co-pilot" replaced a multi-step filtering process (Average of 12 clicks per query) with a single natural language input, significantly lowering the cognitive load for non-technical stakeholders.
• Trend Velocity Accuracy: The "Month-over-Month" performance curves provide a 42.5% more granular view of trend momentum compared to traditional static quarterly reports.

03

Emma's Tree

2023 June - 2024 June · The June 19th Project, Personal Project · Designer & Developer

Overview

• I built this tree for my best friend, a med student who wanted greenery without the maintenance stress. Using biomimicry, the tree reacts to environmental changes just like a real tree, functioning as both interactive companion and a functional monitor that won't die.

Key Highlights

• The "Bloom" Effect, Color Transformation: Temperature Sensitive Filament
• Moisture Sensing: Moisture Sensor
• Visual Feedback: LED Indicators
• Growth & Structure: 3D Pen
• Sustainbility: Solar-powered Battery Bank

Design Thinking

• I wanted to design using biomimicry to solve "care fatigue," so I built a synthetic plant that basically lives and reacts right along with Emma. Using temperature-sensitive filaments and environmental sensors, the tree acts as a responsive sensory anchor that mimics the energy of a real plant.
I used natural materials like real moss and a ceramic pot to keep the tech feeling grounded rather than clinical. Emma was a busy med student, so I wanted to give her those restorative nature vibes in a way that's actually durable and emotionally meaningful.