🇰🇷 한국어

Yeji Kim

Product Engineer

010-2421-1226 heyzhub@gmail.com github.com/heyz-hub linkedin.com/in/heyzhub

I'm an engineer who finds genuine excitement in defining real-world problems and working through how to solve them. I enjoy turning ideas into concrete structures, and I focus on implementing those structures into products that create meaningful changes in user experience. From data pipelines to search infrastructure and agent execution engines, I've worked across the full stack, focused not on delivering information, but on building systems that lead to action. I structure complex data, minimize the need for human intervention, and ship products that change how people actually work.

Experience

Factagora

Factagora

Full-time

Engineering

Jun 2025 – Present (9m)

[AI Agent] Autonomous Execution Engine (ReAct-based Research, Decision & Tool Execution Automation)

Apr 2026 – Present
  • · Built a ReAct-based AI agent execution engine fully automating research → decision → execution
  • · Separated sessions (Prediction/Claim), action loops, and tools for purpose-optimized agent structure
  • · Introduced LLM adapter layer enabling Claude ↔ Azure OpenAI switching (cost/performance optimization)
  • · Designed agent memory·skills system for cross-session long-term memory with learning capability
  • · Integrated search and fact-check features as agent skills, connecting directly to the execution engine
  • · Designed per-token/skill billing system with real-time usage control and session interruption
  • · Applied sliding window context management to optimize response quality vs. cost
  • · Built stable execution environment with async worker (Azure Queue) + Graceful Shutdown
  • · Structured agent execution logs and action history for a debuggable audit system
PythonAnthropic ClaudeAzure OpenAIReActAzure QueuePostgreSQLFastAPIDocker

AI Agent Prediction Competition Platform — Community Feature Development

Feb 2026 – Present
  • · Designed a 3-tier league system (Daily/Weekly/Monthly) and 5-stage tier structure (Rookie → Oracle)
  • · Built a cron-based daily league snapshot pipeline for automated ranking aggregation
  • · Implemented agent score trend graph with tier band overlay using Recharts
  • · Built real-time notification system using Supabase Realtime WebSocket (6 notification types)
  • · Implemented RLS-based notification access control and read-state management
  • · Built email notification infrastructure with Nodemailer + HTML templates and per-user subscription settings UI
Next.jsTypeScriptSupabaseSupabase RealtimePostgreSQLRechartsNodemailer

[AI Search] Search Infra & MCP Integration (Hybrid Search Engine & MCP-based B2B Interface)

Mar 2025 – Present
  • · Built a hybrid search engine combining keyword search and vector similarity
  • · Designed automatic fallback on embedding failure to ensure search availability
  • · Built news data pipeline (DB-first lookup + API fan-out + async ingestion)
  • · Built Knowledge Graph-based content tracking system for source verification
  • · Designed Search / Agent APIs and B2B interface for external service integration
  • · Implemented MCP-based AI tool interface enabling direct feature calls from ChatGPT and Claude
  • · Designed GA4 and PostHog events to visualize the search query → result funnel and quantify bottlenecks
  • · Built CI/CD pipeline (GitHub Actions + Docker + Azure) for automated deployment
PostgreSQLVector DBKnowledge GraphMCPREST APIDockerAzureGitHub Actions

[Frontend SDK] Reusable AI Chat & Graph SDK (Streaming Visualization & Multi-runtime SDK)

Jan 2026 – Feb 2026
  • · Modularized chatbot, streaming, and graph visualization logic into a reusable SDK
  • · Built auto-branching rendering engine for DG/TKG/Evidence graph types
  • · Separated SSE streaming into React Hook + Web API layers for multi-runtime reuse
  • · Optimized bundle size with multi-entry + tree-shaking structure
TypeScriptPythonReactFastAPIAzure OpenAIRAGpgvectorSSESDKTurborepoZustandD3SupabaseDockerTree-shakingOpenTelemetry

[Knowledge Graph] Temporal KG Architecture (Time-range Relationship Modeling & Async Extraction Pipeline)

Dec 2025 – Feb 2026
  • · Designed Temporal Knowledge Graph architecture structuring documents on a time axis
  • · Built async worker-based TKG extraction pipeline for scalable large-volume document processing
  • · Applied LLM-based entity normalization for canonical entity consolidation and data consistency
  • · Modeled relationships using TSTZRANGE for event flow exploration across time
  • · Implemented two-stage search engine combining vector search and relationship traversal
PostgreSQLVector DBKnowledge GraphDockerAzureGitHub Actions

[Chrome Extension] DeepVerify (1-click Verification Flow & Real-time Fact-checking Extension)

Nov 2025 – Present
  • · Led Chrome Extension development from product planning through architecture design
  • · Reduced user friction with a 1-click verification flow, improving usability
  • · Implemented real-time detection logic for SPA environments using MutationObserver
  • · Introduced Clarity and PostHog to analyze heatmaps and session replays, identifying high drop-off points → drove Popup to side panel UI transition
  • · Designed Service Worker-based authentication proxy for security and API stability
  • · Ran hypothesis-driven A/B test on landing page to optimize conversion rate (CVR) by acquisition channel
  • · Implemented real-time claim detection and highlight features, securing demo results
Chrome ExtensionTypeScriptVue.jsMutationObserverService Worker

[Data Pipeline] Live Research Platform (Event-driven Data Collection & Databricks Distributed Pipeline)

Jun 2025 – Oct 2025
  • · Built a plugin-style data pipeline using a collector registry pattern
  • · Designed event-driven structure from collection → processing → storage → post-processing
  • · Built Databricks-based distributed processing pipeline for stable large-scale data handling
  • · Normalized unstructured data to standard Markdown and connected to Knowledge Graph pipeline
  • · Designed LLM-based causal relationship extraction system for high-precision relation data
  • · Implemented SSE-based real-time research interface
  • · Designed Next.js API as authentication proxy for integrated security and communication
  • · Built end-to-end workflow from content creation → editing → saving
ReactNext.jsTypeScriptSSEDatabricksAzure BlobKnowledge GraphLLM

Brainz Company

Full-time

Dev Group 2 – Infrastructure Web Team

Sep 2022 – Sep 2024 (2y)

[Fullstack/Web] Zenius Web Console (IT Infrastructure Monitoring Dashboard & Backend API Development)

Sep 2022 – Sep 2024
  • · Built operator monitoring dashboard UI based on server, network, and system status data
  • · Integrated with Spring-based backend APIs to reliably reflect infrastructure monitoring data on screen
  • · Improved feature stability and maintainability by analyzing and refactoring legacy code
  • · Analyzed and resolved issues found during QA and internal testing to meet production standards
  • · Set up test server and test DB environments during version upgrades and performed DB migrations
  • · Wrote and validated PostgreSQL migration scripts for schema changes
  • · Performed Oracle ↔ PostgreSQL data migrations per client requirements, ensuring data integrity
JavaJavaScriptPostgreSQLSpring

Skills

Overview.

  • I start from real user problems and lead the full product process — from planning through development to deployment.
  • I have end-to-end experience building data pipelines, frontend, server, and infrastructure within a single product.
  • I don't believe the latest technology is always the right choice — I trust that the best option depends on the situation.
  • When the work demands it, I actively explore unfamiliar areas and push for the best outcome regardless.
  • I'm deeply interested in methodologies for getting more out of AI, and I continuously explore and apply new techniques and tools in my day-to-day work.

Communication.

  • I actively share opinions on the product as a whole, and speak up freely regardless of role or seniority.
  • I prefer analytical conversations backed by reasoning, and I don't hesitate to voice my thoughts in any setting.
  • Rather than pushing my own view, I stay open to all perspectives while making sure to surface points worth considering together.

AI / Systems.

▾
Anthropic ClaudeAzure OpenAIReActLLMKnowledge GraphMCP
  • I have experience designing a ReAct-based AI agent execution engine with long-term memory and credit billing for autonomous workflows.
  • I can design Knowledge Graph and Temporal Knowledge Graph architectures.
  • I have experience building MCP servers to connect internal features as native tools in ChatGPT and Claude.
  • I can implement two-stage search combining pgvector-based vector search with graph traversal.

Backend & Infra.

▾
PythonNode.jsFastAPIPostgreSQLpgvectorVector DBRAGDatabricksSpringJavaDockerAzureAzure QueueGitHub Actions
  • I can design and operate hybrid search engines combining PostgreSQL full-text search with vector similarity.
  • I can design REST APIs with FastAPI and build external interfaces for B2B integration.
  • I have experience designing plugin-style data pipelines using a Collector Registry pattern.
  • I have experience building large-scale data processing pipelines in a Databricks distributed environment.
  • I can build and operate CI/CD pipelines using GitHub Actions, Docker, and Azure.
  • I have experience designing async worker structures with Azure Queue and Graceful Shutdown handling.
  • I have experience performing Oracle ↔ PostgreSQL data migrations while ensuring data integrity.

Analytics / Growth.

▾
PostHogGA4Microsoft Clarity
  • I can collect and analyze user behavior data using PostHog, GA4, and Microsoft Clarity to drive product improvements.
  • I have experience designing and running A/B tests to optimize conversion rates (CVR).

Frontend.

▾
ReactNext.jsTypeScriptJavaScriptVue.jsTailwind CSS
  • I can design structured layouts and operate production services using React, Next.js, and TypeScript.
  • I can build clean and consistent UIs using shadcn/ui, Tailwind CSS, and similar libraries.
  • I have experience designing and deploying Chrome Extensions, including SDK bundle optimization with multi-entry and tree-shaking.
  • I can design SSE streaming as a reusable React Hook layer that works across browser, server, and extension runtimes.
  • I have experience building production services with Vue.js.

Education

Hanyang University

Hanyang University

Bachelor's Degree · Physics

Mar 2011 – Aug 2016

Certifications & Related Experience

YDream School 5th

2025.03 – 2025.12

Data Engineering Track

Python-based data pipeline implementation, SQL · PostgreSQL · NoSQL databases, ETL/ELT design, large-scale data processing with Apache Spark

★ Excellence Student · Minister Award

Sparta Coding Club

2022.05 – 2022.09

Cloud Web Developer Training

Cloud and web development fundamentals, backend development

Languages

English

Conversational

Presented products and conducted VC networking in English at the 2025 Plug and Play Summit (Singapore)

Japanese

Elementary

Able to communicate in everyday and travel contexts; basic listening and speaking