Eric Y. Kim^1,2

Real-time economic decision agent that orchestrates Bedrock AgentCore runtimes, live market data, and a Next.js UI to explain how each recommendation is produced. Streams conductor + data-fetcher status events, routes structured subtasks to FRED and Yahoo Finance tools, and persists cross-session memory for consistent guidance.

AI-powered data analysis platform that enables creation of domain-specific custom agents tailored to specialized scientific workflows. Features a Jupyter-like environment with local file access, allowing researchers to build and deploy intelligent agents that understand their specific domain context - from proteomics to drug discovery. Sole architect and developer, focusing on making complex scientific analysis accessible through conversational AI.

Interactive simulation tool helping researchers design optimal protein evolution strategies for CRISPR-Cas9 PAM variants. Compares random vs. ESM-C guided sampling to determine ideal experimental parameters (rounds, samples per round). Provides evidence-based recommendations for resource-efficient directed evolution experiments. Based on fitness models from Kim et al., Nature 2025.

Financial RAG system specializing in Korean regulatory filings (DART/Korean SEC). Built sophisticated PDF and table parsing pipeline to extract structured data from complex financial reports, with LLM-enriched metadata for precise retrieval. Implements advanced PostgreSQL JSONB queries for semantic search across financial tables. Natural language interface in Korean/English enables instant analysis of financial statements, making regulatory data as queryable as a conversation.

Interactive 3D visualization tool for exploring protein fitness landscapes and epistatic interactions between amino acid positions. Analyzes pairwise interactions at key PAM positions (1135, 1136, 1218, 1219, 1335, 1337) with both landscape and interaction modes. Uses experimental data and ML models from Kim et al., Nature 2025.