Kikang Kim

Backend Engineer

I design stable and scalable service architectures.

Seoul, South Korea · Open to relocation

|

About

Backend engineer with 5+ years building event-driven systems and data pipelines, primarily in healthcare and B2B SaaS. Redesigned a clinical data pipeline for 94.8x throughput gain and grew a web analytics platform from 300K to 10M+ monthly pageviews.

Experience

  1. Aug 2024 — Present

    Backend Engineer · AITRICS

    Series C AI healthcare startup — clinical decision support systems

    • ▹REST → GraphQL migration with Strawberry + DataLoader, eliminating N+1 and over-fetching while improving client development velocity
    • ▹Replaced scheduler-based batches with RabbitMQ Pub/Sub per-patient event flow, delivering 94.8x average throughput and real-time alerts
    • ▹Built Retry/DLQ/Sequence + idempotency-key handling for zero-message-loss processing
    • ▹Designed Scoring Manager pipeline — fused AI model predictions with patient chart context to surface clinical decision support data
    • ▹Reworked JWT auth — single-active-token, bcrypt hashing, role-based permissions, AES-encrypted audit log, lockout policy
    • ▹Enforced exclusive locking on patient status history and a minor-patient deletion policy for data integrity and privacy compliance
    • ▹Tracked P50/P95/P99 with Datadog APM + Sentry, wired error-rate, service-health, and on-call alerting
    • ▹Built an AI-agent on-call bot — 117 requests from 21 engineers over 10 weeks, 85.5% resolution rate, 2.6× MoM growth, 20% team MTTR reduction
    ▸AITRICS-VC— Real-time vital sign monitoring and AI prediction system
    • Python
    • FastAPI
    • GraphQL
    • Docker
    • RabbitMQ
    • Datadog
    • Sentry
    • Claude Code SDK
  2. Nov 2020 — Aug 2024

    Backend Engineer · Plan.I

    • ▹Owned end-to-end backend design and operations across VODA, VORY-Q, AI Canvas, and AIVORY (B2B/B2C)
    • ▹Built VODA's behavioral data collection pipeline on AWS API Gateway + SQS, supporting growth from 300K to 10M+ monthly PV
    • ▹Migrated VODA's ingestion pipeline from SQS to on-prem Kafka, cutting infra cost and unlocking higher throughput and replay flexibility
    • ▹Cut VORY-Q recommendation P95 latency by 90% with LRU caching on hot response paths
    • ▹Shipped AIVORY enterprise search — Oracle DB integration plus an automated deployment pipeline that shortened enterprise/government onboarding
    • ▹Ran Kubernetes GPU resources with a Redis workload-aware scheduler, and optimized Elasticsearch incremental indexing to cut visualization time by up to 95%
    ▸VODA— Web analytics platform — user behavior collection and analysis▸VORY-Q— Personalization & recommendation engine▸AI Canvas— AI image generation platform▸AIVORY— Enterprise / government search engine solution
    • Python
    • Django
    • FastAPI
    • AWS
    • Kafka
    • SQS
    • API Gateway
    • Docker
    • Kubernetes
    • Redis
    • Elasticsearch
    • Oracle

Side Projects

  • komaps preview

    komaps

    Trilingual SEO hub for international K-pop and K-drama fans planning pilgrimage trips to Korea — filming locations, idol cafés, and neighborhood guides in English, Japanese, and Simplified Chinese

    • ▹Interactive discovery map with MapLibre GL + OpenFreeMap tiles — zero tile-vendor cost
    • ▹Single-runtime architecture: Next.js 16 RSC + Drizzle on Vercel Fluid Compute, no separate backend service
    • ▹Cloudflare R2 + Sharp image pipeline, DB-backed sitemap with ISR plus secret-protected on-demand revalidate
    • ▹HMAC cookie admin sessions, many-to-many mapping across locations, dramas, and idols
    • Next.js 16
    • React 19
    • TypeScript
    • Drizzle
    • Supabase
    • Cloudflare R2
    • MapLibre GL
    • Vercel

  • recsys-pipeline

    Production-grade recommendation pipeline for 50M DAU commerce — 4-Plane architecture with 3-Tier serving, runs locally with docker-compose, scales to 500K RPS on Kubernetes

    • ▹3-stage serving: Two-Tower candidate generation → session-aware re-ranking → DCN-V2 GPU scoring
    • ▹4-Plane architecture (Data/Stream/Batch/Control) with Envoy, Flink, PySpark, and Airflow
    • ▹Cloud-agnostic: single docker-compose up for local development, Kubernetes manifests for production
    • Go
    • Envoy
    • Apache Flink
    • Redpanda
    • DragonflyDB
    • Milvus
    • PySpark
    • Docker

  • LoL Spell Checker

    League of Legends enemy summoner spell cooldown tracker — zero-dependency single-file HTML deployed to Cloudflare Pages

    • ▹Track 5 enemy champions (Top/JG/Mid/AD/Sup) with 3 spell slots each (180s/240s/Flash)
    • ▹Cooldown reduction toggles: Cosmic Insight (+18 haste), Ionian Boots (+10 haste) auto-applied
    • ▹Real-time countdown + ETA display, game timer, EN/KR language toggle
    • HTML
    • JavaScript
    • CSS
    • Cloudflare Pages

  • auto-card-news

    Automated RSS-to-Threads card news pipeline — fetches news, generates 6-card carousel PNGs, and publishes via Graph API

    • ▹End-to-end pipeline: RSS fetch → Playwright scraping → AI summarization → HTML/CSS card rendering → Threads publish
    • ▹Immutable frozen dataclass models, URL dedup + publish history for duplicate prevention
    • ▹Cloudinary image hosting with automatic cleanup, APScheduler-based periodic execution
    • Python
    • Playwright
    • Jinja2
    • Cloudinary
    • Threads API

Skills

Languages

  • Python
  • SQL

Frameworks

  • FastAPI
  • Django
  • GraphQL
  • REST

Infrastructure

  • Docker
  • Kubernetes
  • Nginx
  • Jenkins
  • CI/CD

Cloud

  • AWS EC2
  • S3
  • RDS
  • SQS
  • API Gateway
  • OpenSearch
  • Lightsail

Data

  • Redis
  • Elasticsearch
  • RabbitMQ
  • Kafka
  • Celery

Monitoring

  • Datadog
  • Sentry
  • Prometheus
  • Grafana

AI-Assisted Engineering

  • Claude Code SDK
  • MCP
  • Multi-agent Orchestration
  • Eval-driven Iteration

Testing

  • pytest
  • TDD

Education

  1. Sep 2022 — Feb 2026

    B.S. in Computer Science and Engineering · Korea Cyber University

    Completed while working full-time (expected graduation Feb 2026)