Backend engineering. Cloud experience.

Problem & context

Modernized core healthcare SaaS backends for reception, reservation, payment, digital charting, patient lookup, and hospital configuration into consistent Kotlin Spring Boot APIs.

Engineering decisions

• Standardized clinical chart and reference flows, including prescriptions, diagnoses, templates, consent forms, edit locks, and delete logs, with OpenAPI and exception documentation.
• Hardened reservation and reception integrations with broad-query guards, linked reservations, department/doctor/date counts, device settings, receipt creation, and patient-app wait-list notifications.

Evidence & outcomes

• Stabilized payment and treatment-room operations through transactional approval/cancel/refund/credit paths, card-terminal history, receivable checks, socket waiting screens, voice guidance, assignment statistics, and rollback paths.

Problem & context

Helped move a healthcare SaaS service fleet from single-service deployment thinking into Kubernetes and ArgoCD-managed application releases.

Engineering decisions

• Worked with Helm charts for patient-facing APIs, hospital admin APIs, automatic reception, payments, waiting displays, procedure allocation, charting, scheduling, interoperability, and shared Spring Boot service templates.
• Standardized Kubernetes templates for deployments, services, ingresses, security groups, HPA, PDB, KEDA ScaledObject, Istio VirtualService, and ServiceMonitor coverage instead of one-off YAML.

Evidence & outcomes

• Supported observability and operations manifests around kube-prometheus, Grafana, Alertmanager, node-exporter, kube-state-metrics, Datadog agent, Jaeger, Kafka/Debezium, Karpenter, and ArgoCD-managed releases.

Problem & context

Built a healthcare messaging backend around Socket.IO/Netty, Kafka events, Redis/Redisson state, JPA/Liquibase persistence, translation, media, tags, reactions, friends, hospitals, and read offsets.

Engineering decisions

• Implemented controllers and services for chats, messages, participants, read state, history, media, reactions, tags, friends, users, and user-hospital unread counts.
• Handled realtime delivery through Socket.IO room subscriptions, chat-message, typing, and read-receipt events, then fanned persisted messages out through Kafka producers and consumers.

Evidence & outcomes

• Added launch-support surfaces around translation configuration, AWS Secrets Manager and S3 integration, Datadog tracing, Slack notifications, request logging, and quiet socket error handling.

Problem & context

Built data and operations pipelines that made product usage and business metrics visible outside the app runtime across patient-app, review, campaign, credit-balance, and hospital operations flows.

Engineering decisions

• Prefect flows collected daily and monthly metrics from Aurora MySQL, BigQuery/GA, Redis, AWS utility modules, and service SQL modules, then published Datadog metrics for dashboards.
• Maintained auto-discovery deployment registration so @flow entrypoints were pushed to the EKS work pool with namespace, service account, image, image tag, and pull-policy configuration.

Evidence & outcomes

• Built backfill, tag-rewrite, metric-history export, and dashboard-update scripts so historical KPI series could be replayed without breaking existing graphs.

Problem & context

Built a multi-tenant B2B platform connecting pharmacy field sales with headquarters and supplier operations: visits, sales, returns, collections, vehicle inventory, purchasing, dispatch, and reporting.

Engineering decisions

• Implemented atomic visit posting, locked stock and balance updates, idempotent completion, signed financial ledgers, and linked reversals instead of rewriting posted history.
• Enforced tenant and ownership scope through JWT validation, service authorization, database connection context, and PostgreSQL row-level policies; separated supplier views from administrator cost and profit data.
• Persisted mobile drafts before versioned submission, retained uncertain payment retries, and distinguished supplier, central-warehouse, and vehicle fulfillment with their own inventory and settlement effects.

Evidence & outcomes

• Delivered React web and Expo/React Native workflows backed by Terraform-managed ECS and private RDS, with separate UAT identities and historical price, cost, tax, and margin snapshots.

Problem & context

Research-first quant pipeline for MES/ES intraday opportunity selection, turning Databento OHLCV and MBP feeds into clean features, labels, deterministic baselines, and model-ready validation gates.

Engineering decisions

• Built the market-data path from raw Databento files into bronze/silver Parquet datasets, strict training matrices, chronological splits, manifests, and reproducible research reports.

Evidence & outcomes

• Designed the research gate around feature information-coefficient reports, strategy-family benchmarks, horizon and roll-rule sensitivity, lookahead audits, MBP microstructure diagnostics, latency/fill/cost stress, and candidate failure attribution.

Problem & context

Cloud orchestration research connecting customer (CSC) service objectives and provider (CSP) resource constraints through eight policy roles, guarded execution, and independent outcome evaluation.

Engineering decisions

• Defined one owner for recovery, provider capacity, admission, service sizing, placement and accepted-workflow dispatch; interference and diagnosis remain advisory roles.
• Built a Referee that checks scope, freshness, capability, SLOs, budgets and conflicts, with a persistent execution journal and verified native Kubernetes postconditions.
• Added optional ridge-regression ranking over feasible proposals, bounded recurring fits, frozen paired promotion evidence and operator rollback; deterministic A–H policies remain the default.

Evidence & outcomes

• Separated customer goodput and provider cost evaluation from proxy rewards and historical Kind comparisons. Gateway deployment, infrastructure permission isolation and measured cloud gains remain research milestones.

Problem & context

Built a native, local-only macOS application that connects live TCP/UDP and Docker evidence to reviewed service intent, ownership-aware lifecycle control, dependency-aware projects, and restorable workspace sessions.

Engineering decisions

• Separated disposable runtime observations from versioned managed-service definitions, then guarded every external action with strong process fingerprints, listener-ownership checks, protected-process rules, and exact Docker/Compose routing.
• Exposed the same application-owned control plane to Codex through an MCP 2025-11-25 STDIO helper and current-user Unix socket, with stable IDs, optimistic revisions, secret-safe results, and single-use destructive previews.

Evidence & outcomes

• Shipped 82 production MCP tools and validated the native app with 210 passing tests; matched Release measurements reduced closed-window CPU from 69.2% to 0.118% over 901 seconds.

Problem & context

Near-startup proptech product that helps founders evaluate Seoul trade areas and shortlist commercial property listings through score-driven analysis, map workflows, and recommendation APIs.

Engineering decisions

• Built backend and frontend surfaces for map-based trade-area analysis, property search, Top 3 recommendations, saved analyses, and detailed listing inspection.
• Implemented API flows around analysis creation, polling and SSE progress, property filtering, structured prompt search, horizon survival scores, score explanations, and metric references.

Evidence & outcomes

• Shaped product UX across Analyze, property listings, Detail, History, shortlist/compare, budget filters, and natural-language property exploration; the project went on to win the 2026 Seoul Big Data Utilization Contest Grand Prize.

Problem & context

A small mobile project for finding travel companions, planning trips, and chatting.

Evidence & outcomes

• Built Kotlin/Spring Boot APIs and React Native flows; received the Excellence Award at Kyung Hee University’s 2026 SemoThon.

Problem & context

Built a Korean-learning product through SWMaestro, covering ideation, planning, backend development, deployment, and app launch.

Engineering decisions

• Developed video-processing flows for subtitle-based learning and speaking practice.
• Used NLP and OpenAI APIs to extract vocabulary examples and usage context from learning material.

Evidence & outcomes

• Owned product execution across backend APIs, Flask services, nginx deployment, and launch preparation.

Problem & context

Turned repeated integration friction into explicit platform patterns across backend, messaging, observability, and deployment repos rather than leaving each issue as a local fix.

Engineering decisions

• Standardized hospital-scoped request context through argument resolvers, Feign auth-token clients, Secrets Manager and Parameter Store configuration, and multi-module application/domain boundaries.
• Improved operational visibility with Swagger/OpenAPI customizers, exception reason docs, request/response logging, Slack error routing, Datadog tracing/metrics, ServiceMonitor coverage, and Prometheus/Grafana manifests.

Evidence & outcomes

• Hardened async and integration paths around SQS, Kafka/Debezium, Redis/Redisson, KEDA autoscaling, PDB/HPA resilience, and transactional facades for payment, reservation, receipt, and allocation flows.

Problem & context

Built a local database lab that connects reproducible concurrency scenarios to a React observability workspace, backed by a Go API and PostgreSQL.

Engineering decisions

• Connected transaction actors, lock inspection, query plans, workload experiments, and saved replays through a shared evidence timeline.
• Designed monitoring views that distinguish current gauges, interval rates, cumulative counters, and unavailable evidence.

Evidence & outcomes

• The local application runs against PostgreSQL 16; its engineering article documents the scenarios and evidence boundaries.