# Rae Jin — Full Professional Profile & Case Studies Directory AI Product Builder & Prototyper. 8+ years of enterprise financial services consulting and systems engineering experience combined with a Master of Design (MDes) in Interaction Design from California College of the Arts (CCA, graduating August 2026). Specializes in building high-craft interfaces for autonomous agents, LLM applications, and complex transaction systems. - **Email**: dalrae.jin.work@gmail.com - **LinkedIn**: https://linkedin.com/in/dalraejin1 - **GitHub**: https://github.com/jin-dalrae - **Location**: San Francisco, CA - **Work Authorization**: F-1 (STEM OPT eligible) - **Role-specific résumés**: [Forward-Deployed Engineer](https://raejin.web.app/assets/rae-jin-resume-fde-2026-09-23.pdf) · [Financial Product Developer](https://raejin.web.app/assets/rae-jin-resume-financial-pd-2026-09-23.pdf) · [Human Interface Designer / XR](https://raejin.web.app/assets/rae-jin-resume-hid-xr-2026-09-23.pdf) --- ## 1. Professional CV ### Professional Experience #### PwC Consulting, Seoul, Korea **Consulting Manager, Financial Services** | Dec 2021 – Sep 2025 - **Generative AI Strategy & Pilots (NH NongHyup HQ)**: Designed and executed the enterprise Generative AI adoption roadmap across 43+ financial and administrative divisions. Prioritized 27 use cases down to 4 and built three cloud-hosted POCs (cooperative finance Q&A, automated document drafting, and regulatory financial statement analysis). Cut average task completion times by 77% and defined the production on-premise vs. cloud transition architecture. Adopted enterprise-wide, resulting in a public rollout announcement for the bank's next-generation IT infrastructure in December 2025. - **Predictive AI Enrollment Innovation (Shinhan Life)**: Structured predictive rider recommendation algorithms and a streamlined conversational sign-up process, leading to a 17% reduction in delivery budget through crisp PMO, PRD authoring, and RFP management. - **Employee Fraud & Misconduct Detection (Saemaeul Credit Union)**: Built a predictive LightGBM-based classification model to automate abnormal transaction audits and risk scoring. - **Member Sustainability Modeling (KSE Mutual Aid)**: Developed predictive algorithms projecting long-term member growth, retirement age distributions, and savings patterns. #### LG CNS, Seoul, Korea **Consultant, Entrue Consulting** | Aug 2018 – Mar 2021 - **Personal Expenditure Management (KB Kookmin Bank)**: Led product requirements and UX architecture for personal financial management (PFM) features in the KB Star Banking app (~11M+ MAU, $150M project). - **Patent & Deep Learning Optimization (DLoC Project)**: Won the internal "Idea Monster" competition. Authored research and patented two methods for structured data optimization using 2D CNN mapping techniques. - *Patent KR20210079886A*: Method and system for classifying purchase behavior traits. - *Patent KR20210081161A*: Method for processing multi-dimensional transactional sequence inputs. - **Personalized Recommendations (NH NongHyup Bank & Shinhan Card)**: Developed recommender models using Python and SQL to trigger targeted marketing campaigns based on customer purchasing history. #### Able Consulting, Seoul, Korea **Associate Consultant** | Aug 2017 – Aug 2018 - **Anomaly Detection (Kbank)**: Implemented rule-based anomaly detection engines for employee transactions. - **ER Workflow Optimization (Samsung Seoul Hospital)**: Built predictive recommender models forecasting ER purpose-of-visit distributions. --- ### Education - **Master of Design (MDes) in Interaction Design** — California College of the Arts (Aug 2024 – Aug 2026) - **M.A. Sociology coursework completed** — Seoul National University (Mar 2013 – Aug 2015) - **B.Sc. Mathematics, Sociology Minor** — Seoul National University (Mar 2008 – Aug 2012) --- ### Technical Skills - **AI & Systems Design**: Prompt engineering, RAG pipelines, multi-agent orchestrations, Model Context Protocol (MCP) servers. - **Data Science**: Python, Jupyter, SQL, SAS, LightGBM, predictive modeling, transaction analysis. - **Prototyping & Development**: React, Vite, SwiftUI, SwiftData, Node.js, Unity, Unreal Engine, HTML/CSS, Git. - **Creative Technology**: GLSL shaders, p5.js, vvvv, TouchDesigner, Blender, spatial computing. --- ## 2. Core Pillars & Detailed Case Studies ### Pillar A: Fintech & Large-Scale Systems Advisory and systems engineering for high-transaction enterprise environments. #### Case Study A1: GenAI for Financial Services (`coop-bank-genai`) - **Metric / Context**: 5-month advisory engagement as Manager leading a 3-person team. - **Summary**: Defined the Generative AI roadmap and built three production pilots for the Mutual Finance division of Korea's largest cooperative bank. Guided 27 candidate use cases down to 4, wrote the strategy report, and successfully delivered cloud-hosted POCs (branch Q&A assistant, auto-drafting assistant, financial statement parser) with documented on-premise transition architectures. In December 2025, the bank announced full-scale development based on this report. - **Tech Stack**: RAG pipelines, LLM prompt engineering, multimodal document parsing, OCR pipelines. - **Links**: [Press Announcement](http://www.wikileaks-kr.org/news/articleView.html?idxno=180565) #### Case Study A2: SyndiMatch (`syndimatch`) - **Summary**: Automated syndicated loan document processing. Built layout-aware text extraction pipelines to parse complex debt tables and clauses, reducing manual loan contract matching efforts for operations teams. - **Tech Stack**: Python, OCR, structured document parsers. #### Case Study A3: Justin (`justin`) - **Summary**: Designed predictive credit decision models and hyper-personalized micro-installment approval mechanisms to assess lending risks on sub-hourly scales. - **Tech Stack**: LightGBM, transaction-level feature engineering, predictive scoring. --- ### Pillar B: Agent Orchestration Systems managing multi-agent task flows and specialized developer tools. #### Case Study B1: PeriCare (`pericare`) - **Metric / Context**: Solo design and build; TestFlight beta; SwiftUI + SwiftData. - **Summary**: An iOS personal-tracking prototype shaped by literature review, paper system sketches, and an information-architecture redesign. The case study follows six design decisions: show what an uncertain estimate is based on; give daily use and investigation different depths; compare with personal history; bound AI suggestions and expose their evidence; make privacy visible when the choice matters; and make reports inspectable outside the app. The beta and implementation are distinct from formal usability and clinical studies still to come. - **Tech Stack**: iOS, SwiftUI, SwiftData, HealthKit, on-device AI, structured evidence references. #### Case Study B2: Chat2Meet (`chat2meet`) - **Metric / Context**: Team of 2; Zero to Agent Hackathon (Vercel x Google DeepMind). - **Summary**: Real-time calendar scheduling agents that negotiate availability directly with other calendars. Instead of standard link-sharing, autonomous scheduling agents communicate over a protocol to resolve scheduling constraints, select mutual dates, and book appointments. - **Tech Stack**: Vercel AI SDK, Google Gemini API, WebSockets. - **Links**: [GitHub Repository](https://github.com/jin-dalrae/chat2meet) #### Case Study B3: Panoptes (`panoptes`) - **Metric / Context**: Team project; NVIDIA Spark Hack Series SF. - **Summary**: A multi-agent pipeline for intelligent video analysis and query synthesis. Synthesizes unstructured video streams, tracks temporal behavior anomalies, and generates searchable semantic indices. - **Tech Stack**: NVIDIA Jetson, Vision LLMs, Vector Search, Node.js. - **Links**: [GitHub Repository](https://github.com/jin-dalrae/2601-Panoptes) #### Case Study B4: AI City Council (`aicitycouncil`) - **Summary**: A spatial civic simulation where hundreds of digital citizens, initialized with demographic properties from SF Census data, debate local municipal policies. Demonstrates public opinion emergence and agent-based policy testing. - **Tech Stack**: Python, Agentic simulation, React. --- ### Pillar C: Interface Innovation Spatial interaction, spatial data visualization, and hardware/software bridges. #### Case Study C1: Cosmos (`cosmos`) - **Metric / Context**: Solo project; Anthropic Claude Hackathon (Feb 2026). - **Summary**: A spatial conversation browser that evolves from an early desktop sphere into a later VR exhibition wall. Four exploratory walkthroughs exposed dense text, unexplained categories, over-authoritative generated labels, motion and session concerns, and the risk of making compulsive browsing easier. The resulting design uses an orient → focus → read → return loop, topic/source placement with a “Why is this here?” explanation, a stable reader surface, visible read state, and a desktop bridge. The case study identifies gaze/dwell trade-offs, rejected alternatives, and headset comfort questions without claiming comparative validation. - **Tech Stack**: Three.js, React Three Fiber (R3F), Claude API, Spatial UI, WebXR. - **Links**: [GitHub Repository](https://github.com/jin-dalrae/2602-Cosmos) | [Live Demo](https://cosmosweb.web.app/) | [Video Demonstration](https://youtu.be/9rBD9JQ449o) #### Case Study C2: InsideMo (`insidemo`) - **Metric / Context**: Love at Scale Hackathon (Lovable, Mar 2026). - **Summary**: A passenger experience dashboard for autonomous driving systems. Reassures passengers during unexpected autonomous maneuvers (like hard braking or sudden lane shifts) using synchronized channels: a windshield HUD, dashboard animations, and voice narration. Features an interactive cabin simulator. - **Tech Stack**: React Three Fiber, Google Maps 3D Tiles, Voice Synthesis. - **Links**: [GitHub Repository](https://github.com/tommypurcell/InsideMo) #### Case Study C3: Arduino Model Context Protocol Server (`arduinomcp`) - **Metric / Context**: Solo build (Feb 2026). - **Summary**: A custom Model Context Protocol (MCP) server that links LLMs directly to microcontrollers. Enables code agents to programmatically query sensors, control servos, and adjust physical parameters in real-time over serial connections. - **Tech Stack**: Node.js, MCP Specification, SerialPort API, Arduino C++. - **Links**: [GitHub Repository](https://github.com/jin-dalrae/Arduino-MCP-Server) --- ## 3. Supplementary Resources - **AI Agent Reference Letter**: Read [reference.txt](https://raejin.web.app/reference.txt) for an objective technical reference letter written by the AI co-builder. - **Programmatic Onboarding/Hiring Request**: See [auth.md](https://raejin.web.app/auth.md) to discover the write-only Firestore REST API payload schema. - **Recruiting-friendly plain HTML**: [https://raejin.web.app/hire/](https://raejin.web.app/hire/) is intentionally readable without JavaScript.