Hyeonjun Kil
Backend Engineer / Platform Part Leader
Professional Summary
Backend engineer with 5 years of experience building a shared platform for 6 live games (12 branches), including multi-tenant chat that held 20,002 CCU for 900s in a dev load test. My background spans DB optimization, concurrency, and legacy migration across education, healthcare, clinical trials, and gaming. As hands-on lead of a 3-person BE/FE/AI Platform Part, I manage shared done criteria and quarterly OKRs, and verify AI-assisted output with TDD, integration tests, and review.
Experience
GameDuo
2025.01 — PresentMobile game development and publishing company. Shared backend platform design for 6 live game environments (12 operating branches), data pipeline engineering, and common package system operations
- •Single-shard chat bottleneck at 6,000 msg/s: ack p95 767ms, 99.39% acceptance — Personally led sharding and load validation. Separately set shared done criteria and verification templates for the 3-person BE/FE/AI part, tracking work in quarterly OKRs with capacity targets (6,000 CCU, 18,000 RPS, 90,000 RPS burst), Load-test ack p95 767ms→6.6ms, scaling 1→4 shards
- •Archive backfill saturated RDS CPU at 99.8%, throughput 9.3 items/min — Performance Insights isolated a query causing 99.5% of load; replaced N+1, DATE() predicates that prevented index use, and full-column reads with IN() batching, direct date comparison, and selected columns, Backfill throughput up 26x (9.3→247 items/min), Lambda p50 63.2s→1.36s, 15,333 completed with 0 pending
- •Company-wide AI agents needed centralized per-user policies, device identity, and usage tracking — Owned design, operations, and merge approval for device registration, policy, remote-control, and usage tooling; AI agents implemented it before operational handover, Assigned the usage-reporting switch to 92 people on the active roster; all 87 reporting online devices applied it, and 525 automated security tests passed
- •Coding standards, domain knowledge, and verification rules were scattered across individual setups — Unified rules, agents, hooks, and domain knowledge in Git with local RAG search, Every member shares the same AI working context via git pull, with pre-review criteria captured in an automated checklist
- •High DB and CPU load during metrics sync — Evaluated partitioning: write IOPS fell, but read IOPS rose and CPU load persisted, Chose lower Lambda concurrency over a structural change with little benefit for its code complexity
- •Batch processing collection limits (60 days) — Replaced batch runs with EDA on-demand processing to distribute peak load, Expanded metrics range 6x (60→360 days), processing time 2h→5min
- •Code duplication and convention inconsistency across 7 projects — Introduced shared NestJS packages and auto-deploy (10 utility modules + Game Server Kit), Upgrade rollout across 7 projects 3hrs→15min via automation, with shared-code changes consolidated into one package
- •Slow marketing queries across Google Ads/Meta/TikTok data — Normalized read path + index tuning to avoid full-scan on 100-column table, Performance improved 97% (18s→0.5s)
- •Regulatory risk from game probability disclosure requirements — Integrated a DynamicModule probability package across games and built a CDK audit-log pipeline, Achieved 94%+ integration test coverage (12 suites, 115 tests)
Projects: Real-Time Game Chat Server (Multi-Tenant), Marketing Integrated Platform, Internal Common Library System
Dev2 Team Docs Squad Backend Engineer
2023.06 — 2024.08JNPMedi [Detail]
Clinical trial data management solution startup with KRW 16B cumulative investment (MSA-based)
- •Intermittent automatic e-signature failures caused by database deadlocks — Reproduced foreign-key (FK) shared/exclusive-lock deadlocks using SQL and lock tables, Derived two remedies: transaction ordering and SELECT FOR UPDATE pre-acquisition
- •VDR project schedule delay under MVP launch pressure — Emergency sprint with 2 BE + 1 FE: Event-Driven PDF conversion + permission-based workflow, Launched the MVP in 3 weeks and ended synchronous large-PDF conversion timeouts
- •Missing delivery/bounce tracking left email failures visible only after the fact — Received SES Configuration Set events through SNS and logged results in Datadog, Built Slack alerts for delivery failures and shared the implementation through a tech blog and internal seminar
Projects: Maven Docs System Reliability Improvement, Maven VDR MVP Development
Dev Team Backend Engineer
2022.04 — 2023.06Medgo [Detail]
Telemedicine and medication delivery platform startup with 300K cumulative downloads
- •Auth, appointment, and notification changes risked regressions in a single Express app.js — 3-stage progressive migration (single-file→layered→NestJS) + JS→TS, Reached 80% test coverage and established module/layer boundaries and team coding conventions
- •Manual prescription entry bottleneck (3-5min per case) — Naver Clova OCR + MFDS (Korea FDA) API integration for full workflow automation, Pharmacist processing time 3–5min→30sec per prescription
- •Status reflection delay in consultation/dispensing/delivery (tens of seconds) — Socket.IO real-time sync + Web Push notifications, Reduced to under 1 second, eliminated patient wait complaints
Projects: Doctor/Pharmacist Web Service Enhancement, Backend Architecture Migration
Dev Team Backend Engineer
2021.07 — 2022.03SimpleHAN [Detail]
Custom software consulting firm
- •Manual learning management (attendance, completion, progress tracking) — Built LMS full-stack (QR attendance, PDF certificates, Excel bulk registration), Digitized operations for 300+ students per semester
- •Adding matching and counseling to legacy PHP risked regressions — Integrated an independent Spring Boot app via an iframe/postMessage boundary with weighted-average matching, Launched the JOB Agent service while isolating the legacy-system impact surface
Projects: Learning Management System (LMS) Development
Side Projects
Platform / Tooling Engineering
2026.04 — PresentPersonal Projects (Side) [Detail]
Self-directed platform/tooling engineering outside work — development workflow automation framework and side projects
Projects: Public AI CLI Runtime Launcher
Core Competencies
DB Optimization, Concurrency & Distributed Systems
- •Query-plan and lock-contention analysis; Redis distributed locks, Streams WAL, and sharding
Serverless Architecture & Cost Optimization
- •AWS serverless and event pipelines; BigQuery Storage Read API migration
Legacy Modernization & Reliability
- •Incremental Express-to-NestJS migration, TypeScript, shared packages, and test-backed changes
Engineering Productivity & Quality Verification
- •Shared team context, review checklists, and Git hooks for AI-assisted output
Presentations & Writing
Open Source
Technical Skills
Education
Inha Technical College
Bachelor's (Advanced Major Program, Concurrent Employment) • Computer Information Engineering
2022.02 — 2023.02
Inha Technical College
Associate (Concurrent Employment) • Computer Information
2017.02 — 2022.02
Certifications
SQLD (Structured Query Language Developer)
Korea Data Agency • 2025.12
Engineer Information Processing
Human Resources Development Service of Korea • 2023.06
Professional Development
System Design Interview Study [GitHub] — System Design Interview: An Insider's Guide Vol. 1 & 2
2025.08 ~ 2026.04