Skip to content
Jaeyong Kim (Jay) - AI-native Engineer & Systems Architect Profile Photo

Jaeyong Kim (Jay)

AI-native Engineer & Systems Architect

Professional Summary

AI-native engineer and systems architect with 11+ years of experience turning business transitions into executable engineering systems, from 0-to-1 startup builds and post-M&A technology integration to payment and settlement stabilization and Big 4 tax platform development. Currently building a next-generation AI-native accounting and tax platform at EY Han Young, designing a multi-agent tax AI assistant; agentic workflows for investment tax review, VAT reconciliation, and contract review; an Azure serverless runtime; an agentic retrieval knowledge layer; and evidence tracking with expert validation. Previously led Pluglink's carve-out migration stream to exit the TSA after its acquisition of Hanwha Solutions' EV charging business; reduced Pierrot Company's batch payment processing time from 3 hours to roughly 10 minutes through an idempotent asynchronous payment architecture and recovered approximately KRW 330M in unpaid receivables; and improved charger hardware cost and recurring connectivity cost structures at Chain Lightning Company by designing the SaaG (Smartphone-as-a-Gateway) protocol. Also led AI-native product pivots from EV charging data-driven smart charging and pricing optimization to the GPT-based legal work copilot D.PILOT, Documate document AI, Tripmate TravelGPT, AI itinerary planning, and travel companion platforms. Known for converting domain experts' tacit knowledge into data models, retrieval strategies, agent instructions, tool calling, function calling, evaluation criteria, and production workflows, then connecting technical decisions to cost reduction, revenue recovery, operational stability, and organizational adoption. Focused on engineering leadership that designs and implements AI-native products, AI-native teams, and AI-native development lifecycles together.

Key Achievements

AI-native Engineering

Built a multi-agent tax AI assistant for actual tax-practice use at EY Han Young, and designed the engineering structure for converting expert work such as investment tax review, VAT reconciliation, and contract review into AI workflows.

Established the foundation for a core work platform for Big 4 tax professionals (Azure Serverless, Multi-agent System, Vector Search, Hybrid RAG)

Agentic AI Runtime

Defined a traceable agentic retrieval knowledge layer and validation loop by connecting source data normalization, vectorization, source lineage, citation validation, expert review, and approval flows.

Improved verifiability and reusability of tax AI answers (Evidence Tracking, Human-in-the-loop, Risk-tiered Autonomy)

Generative AI Product Pivots

Led product pivots from an EV charging platform into GPT-based legal and document work automation (D.PILOT, Documate), Tripmate TravelGPT, AI itinerary planning, and travel companion platforms, including Azure OpenAI prototypes and user validation flows.

Expanded business opportunities into legal tech, document automation, travel tech, and community domains (GPT-4, Azure OpenAI, Function Calling, Mobile App)

M&A Technology Integration (PMI)

Led the carve-out migration stream required to exit the TSA after Pluglink acquired Hanwha Solutions' EV charging business, and executed a phased serverless transition based on API Gateway and AWS Lambda.

Contributed to technical independence and TSA termination (AWS Lambda, API Gateway, Serverless Framework)

Fintech System Stabilization

Analyzed payment bottlenecks and introduced a Django + Celery asynchronous queue with idempotency design, reducing batch payment processing time from 3 hours to roughly 10 minutes.

Recovered approximately KRW 330M in unpaid receivables and stabilized payment processing (Django, Redis, Celery)

Platform Engineering

Introduced Terraform-based IaC, ArgoCD ApplicationSet, Helm standardization, and an EKS operating model based on Karpenter and Spot instances.

Standardized deployments and improved infrastructure cost structure (EKS, Terraform, ArgoCD)

IoT Cost Innovation

Designed the SaaG protocol based on a BLE-HTTPS bridge, removing LTE modems from chargers and turning smartphones into gateways.

Improved hardware cost structure and eliminated recurring connectivity fees (BLE, Merkle Tree, Python)

Security & Integrity

Designed mechanisms for data integrity verification even in unstable network environments. Designed wallet, point, and coupon consistency models and the KECO public data pipeline, and reliably managed data inconsistency cases during operations.

Strengthened trust in payment, settlement, and charging data (Cryptography, PostGIS, Data Pipeline)

Core Skills

AI-native Engineering & Agentic AI

multi-agent assistant, agent orchestration, tool calling, function calling, domain-specific prompt orchestration, tax workflow automation, agentic retrieval knowledge layer, evidence and citation validation, human-in-the-loop validation, evaluation criteria, AI-native product engineering.

Enterprise IT PM & Business Analysis

requirements analysis, WBS planning, business workflow analysis, acceptance criteria definition, UAT and launch management, coordination across SMEs, development teams, and external vendors.

Cloud & Networking

multi-account environment governance, heterogeneous infrastructure integration through VPN routing optimization, hybrid cloud network design.

System Architecture

enterprise architecture, ERD, interface definition, API design, progressive monolith decomposition, serverless carve-out, distributed transaction management, idempotent payment systems, high-performance asynchronous messaging architecture.

Cloud & Platform Engineering

Azure serverless architecture, AWS Lambda, API Gateway, EKS, ECS on Fargate, Terraform, Helm, ArgoCD, Karpenter, GitHub Actions, Jenkins, AWS EKS-based GitOps pipeline modernization.

Backend Engineering

Python (Django, Django Ninja, DRF), Java (Spring Boot), Node.js and TypeScript-based high-availability API design, Celery, Redis, PostgreSQL and PostGIS, DynamoDB, SNS and SQS, internal common module distribution systems for NPM and PIP.

Domain Experience

tax and accounting AI platforms, legal tech and document AI, travel tech AI, fintech payments, settlement, wallets, EV charging, OCPP, IoT, smart factory MES and LTS, AI education platforms.

Leadership

AI-native team, product, and development lifecycle design; engineering culture and technical standards; startup technology leadership and co-founder experience; M&A technology integration; technology roadmap planning; contribution to government R&D (TIPS) and MS Launcher program selection, securing up to KRW 500M in Azure credits; cross-functional team leadership.

Work Summary

Work Experience

EY Han Young — Assistant Director, IT PM (AI-native Engineering)

2026.03 - Present, professional services in assurance, consulting, tax, strategy, and transactions; founded in 1982; Korean member firm of the EY global network; approximately KRW 800B in Korea revenue with approximately 3,300 employees; approximately USD 55B in global revenue with approximately 400,000 employees

Build a next-generation AI-native accounting and tax platform embedded in the daily work of tax professionals, and establish the product structure, agentic workflows, serverless runtime, and verifiable knowledge layer needed to expand it into an integrated tax AI work platform connecting work-context search, judgment support, review and approval, and knowledge accumulation.
  • Multi-agent tax intelligence assistant engineering: Built an AI assistant that allows tax professionals to ask natural-language questions about work context while multiple agents divide responsibilities for search, judgment, review, and response. Shaped agent role decomposition, retrieval flow, and response validation loops into a usable tax-work product.
  • Agentic tax workflow design and implementation: Designed the structure for converting specialized tax and accounting work such as investment tax review, VAT reconciliation, and contract review into agent-based workflows. Defined business rules, review criteria, user approval flows, and failure and exception handling standards as part of the workflow design.
  • Azure serverless AI runtime design: Designed the runtime environment for operating agentic AI products on Azure-based serverless architecture, including authentication, APIs, functions, storage, orchestration, observability, and deployment boundaries.
  • Verifiable AI knowledge architecture: Defined source lineage, citation validation, evidence tracking, risk-tiered autonomy, and human sign-off workflows so AI agents can trace the basis of their answers and experts can validate them.
  • Agentic retrieval knowledge layer engineering: Normalized, chunked, and vectorized source data, then defined metadata schemas so AI agents can understand and reuse accurate tax context. Modeled relationships across documents, transactions, and review histories, converting scattered work materials into a searchable, verifiable, and reusable knowledge structure.
  • Tax AI Operating System product direction: Established the direction for an integrated tax AI platform where the starting point of expert work, review flows, deliverable generation, approval, and knowledge accumulation connect in a single work layer, beyond standalone assistants or agents.
  • AI-native development lifecycle definition: Defined a development lifecycle for AI products that connects requirements discovery, prompt and agent design, data preparation, evaluation criteria, human-in-the-loop validation, deployment, and monitoring.
  • Coordination across tax SMEs, development teams, and external vendors: Translated tax experts' tacit knowledge into agent instructions, data models, retrieval strategies, and validation criteria, while coordinating implementation scope, quality standards, schedules, and risks with internal development teams and external vendors.
Azure Serverless Architecture Multi-agent System Agent Orchestration Vector Search Hybrid RAG Source Lineage Citation Validation Human-in-the-loop Validation AI-native Development Lifecycle Business Analysis System Architecture UX and UI Planning UAT Vendor Coordination Risk and Issue Management.

Pluglink — Tech Lead

2025.03 - 2025.08, EV charging platform, founded in 2021, private company, approximately KRW 20B in revenue, 30 employees

Integrate the heterogeneous IT infrastructure of the EV charging business acquired from Hanwha Solutions and exit a high-cost TSA (Transitional Service Agreement) within a short timeframe to secure operational efficiency and technical independence.
  • Successful TSA exit through phased serverless migration: Placed API Gateway in front of the legacy monolith to control traffic and progressively separated key functions into AWS Lambda-based serverless services. Led the core technical migration plan and carve-out stream targeting TSA termination, contributing to the transfer of operations and technical independence within the planned timeline.
  • Complex network environment optimization: Resolved connectivity issues among public institutions (KECO), the existing business unit, and the acquired business unit by unifying VPN routing and cleaning up the network topology, eliminating conflicts and securing stable integration.
  • OCPP messaging bridge layer operation and stabilization: Operated and managed a WebSocket-based OCPP messaging pipeline, and reviewed and improved the bridge layer's message transformation and routing logic and DynamoDB-based connection state management flow. Separated and reorganized the SNS and SQS-based message pipeline by domain to ensure consistent processing and operational stability under high traffic.
  • Engineering standardization and operational improvement: Operated and improved a common-module and DTO reuse framework based on an internal NPM registry, increasing development productivity. Standardized and managed Jenkins, GitHub Actions, Slack notifications, and EKS manifests to strengthen deployment consistency and automation.
  • Development culture and process improvement: Moved the development workflow from Bitbucket to GitHub and introduced a code review culture and Git Flow-based collaboration rules, standardizing the mixed development culture after the acquisition.
AWS Lambda Serverless Framework API Gateway Java (Spring Boot) Node.js TypeScript Python DynamoDB SNS SQS RDS Elasticsearch Datadog Jenkins GitHub Actions.
Reason for Leaving: After completing the PMI and initial stabilization mission, the organization shifted into a maintenance phase. Moved to a new challenge where more advanced architecture design capabilities could be applied.

Pierrot Company — CTO

2024.05 - 2025.03, phoneGO electronics subscription service, founded in 2019, private company, approximately KRW 2B in revenue, 10 employees

Resolve revenue leakage caused by payment failures, design a technology roadmap for selection into the government-supported TIPS program, and secure business scalability through architecture modernization.
  • Idempotency-based bulk payment system refactoring: Analyzed bottlenecks and failure rates in the existing synchronous payment process and redesigned it as a Django and Celery-based asynchronous queue architecture. Applied idempotency keys and exponential backoff (Smart Retry), reducing batch payment processing time from 3 hours to roughly 10 minutes, preventing duplicate payments, and building the foundation for revenue recovery.
  • Systematized unpaid receivables recovery process: Planned, developed, and launched a feature that allowed customers to directly view and pay overdue balances. Coordinated product planning, design, frontend, and backend teams to ship it in roughly two weeks, recovering approximately KRW 330M in unpaid receivables as of February 2025.
  • Deployment modernization with IaC and GitOps: Codified infrastructure components such as VPC, EKS, ElastiCache, and IAM with Terraform to improve reproducibility, and introduced ArgoCD ApplicationSet and Helm standardization to standardize the deployment structure.
  • Infrastructure cost and operational efficiency improvement: Improved the infrastructure cost structure through EKS node configuration based on Karpenter and Spot instances, and strengthened operational stability by standardizing resource requests and deployment specifications.
  • CS operations automation and customer feature improvements: Released features that reduced operational bottlenecks, including Kakao AlimTalk notification automation, payment method management, contract lookup, and KCB basic credit information lookup, reducing manual work for the CS organization and improving operational accuracy.
  • R&D strategy and funding acquisition: Led the technology roadmap for an AI engine to forecast used electronics rental demand, contributing to selection for the TIPS program and securing government R&D funding.
Python (Django / DRF) Celery Redis React TypeScript Kubernetes (EKS) Terraform Helm ArgoCD GitHub Actions PostgreSQL Datadog Prometheus Karpenter.
Reason for Leaving: After completing core missions such as TIPS selection and payment system stabilization, accepted Pluglink's recruitment offer to work with larger-scale traffic in the EV charging platform domain.

Chain Lightning Company — CTO & Co-founder

2021.11 - 2024.05, Sparkle EV charging and mobility super app, founded in 2021, private company, early revenue stage, 10 employees

Build an EV charging platform from 0 to 1, use technology to address hardware cost and communication fee burdens, and secure a sustainable charging, payment, and mobility platform model. Later, pursue product pivots into legal work assistance, travel planning, and travel companion services using generative AI.
  • SaaG (Smartphone-as-a-Gateway) protocol design: Designed a BLE-HTTPS bridge that uses smartphones as gateways, removing LTE modems from chargers and creating a structure to improve hardware cost and eliminate recurring connectivity fees.
  • Offline-first security and integrity architecture: Designed a Rolling Key encryption and Merkle Tree-based integrity verification protocol for unstable network environments, securing the consistency and security of charging data.
  • EV charging public data pipeline and recommendation engine: Collected and reconciled KECO public data through a multithreaded pipeline, and implemented route optimization and charger recommendation algorithms based on PostGIS. Designed a recommendation structure that calculates route reachability, expected charging time and cost, and available charging slots.
  • AI smart charging and pricing optimization design: Designed data-driven value-added service directions such as automatic charging scheduling, demand forecasting, alternative charger recommendations, and lowest-cost prediction using user charging patterns, off-peak time windows, and charging history data.
  • Payment, wallet, and coupon consistency model: Designed Bootpay-based payment methods, Deposit and Withdraw wallet models, and coupon and point flows to manage consistency across payments, coupons, and points.
  • Real-time chat and mobile service operations: Built a real-time chat system with Django Channels, Redis-based PubSub, and MongoDB storage, and operated a Flutter-based app, AWS ECS on Fargate, and GitHub Actions deployment pipeline.
  • D.PILOT legal tech and document AI product development: Led development of the GPT-based legal work copilot D.PILOT after 2023. Planned it as a generative AI work assistant for lawyers' document drafting, consultation responses, and office operations, and implemented domain document automation PoCs including Documate contract and IR translation workflows with lawyer review, legal job postings, cover letters, and resume generation tools.
  • Tripmate AI travel and companion platform pivot: Led Tripmate development, including AI-generated travel planning, TravelGPT Q&A, curation of popular places and restaurants, nearby travel companion matching, and community features that connect travelers through post-trip meetups. Enhanced user questions with actual place search results through Azure OpenAI function calling and Google Places integration, and implemented POI recommendation and map prototypes combining Google, Kakao, Naver, and Elasticsearch data. Led an eight-person team to launch the MVP in two weeks.
  • Technology management and external validation: Secured up to KRW 500M in Azure credits through selection for the MS Launcher program, and contributed to selection or recognition in multiple programs including the Pre-Startup Package through Seoul Center for Creative Economy & Innovation with top completion honors, Korea Credit Guarantee Fund NEST 12th cohort, D.CAMP D.DAY, and K-Global SNU Unicorn.
  • Full-stack product development and operations: Designed and implemented login, payment, coupon, wallet, real-time chat, notification, and file management modules. Built CI/CD with GitHub Actions, Docker Compose, and AWS ECS on Fargate deployment pipelines, and operated chat systems based on MongoDB, Redis PubSub, and Django Channels.
Python (Django Ninja / Flask / Streamlit) Flutter AWS ECS on Fargate GitHub Actions PostgreSQL (PostGIS) MongoDB Redis Redis-based PubSub Django Channels Elasticsearch Google Places API Azure OpenAI GPT-4 New Relic Firebase Microsoft Azure.
Reason for Leaving: Built products from 0 to 1 as a founder and carried out several pivots, but closed the business due to market conditions. After organizing the technical assets, accepted Pierrot Company's CTO offer to apply these capabilities in a professional organization.

Additional Experience

Education

Military Service

Certifications

Languages

Selected Activities

Interested in working together?

I'm always open to discussing new opportunities and collaborations. Feel free to reach out.

Get in Touch