I build full-stack applications, backend/data platforms, automation systems, and applied AI tooling with a strong focus on reliability, observability, and real operational problems.
My primary languages are JavaScript / TypeScript / Python. I also have hands-on experience with Node.js, React, Vue.js, FastAPI, PostgreSQL, MySQL, Redis, Docker, SQL, and GitHub Actions.
My background spans software development, fintech, security, digital forensics, fraud/risk analytics, and automation. That domain depth is useful because I do not approach software as isolated code — I am used to translating ambiguous business and operational problems into working systems.
The projects below are portfolio/reference implementations. Data-driven demos use synthetic data, with scope and limitations documented in each repository.
A full-stack evidence-review workspace with source-backed records, role/tenant boundaries, leased SQL jobs, version-checked human approval, and append-only audit history. A separate RAG Lab demonstrates TF-IDF retrieval, citation validation, and an optional OpenAI reasoner.
React · TypeScript · FastAPI · SQLAlchemy · PostgreSQL / SQLite · pytest · Playwright
The public demo uses SQLite and fictional personas. PostgreSQL row-level isolation and concurrency are covered by a separate integration suite.
Live demo · Architecture · Tests · AI boundaries
An interactive usage-analytics and recommendation app with cohort/period filters, adjustable hybrid-ranking weights, explainable score contributions, cold-start handling, and time-split offline evaluation.
React · TypeScript · Vite · Node.js · Playwright · GitHub Actions
The browser and evaluation scripts share the same domain engine. It runs on synthetic sessions without bank-account connections; reported metrics validate this bounded dataset, not real-market performance.
Live demo · Architecture · Evaluation & limitations · Browser verification
An interactive investigation console for directional hop tracing, transaction-graph exploration, and explainable fan-in, fan-out, peel-chain, and risk-proximity signals. Evidence panels connect each result to its underlying synthetic transactions.
React · TypeScript · Cytoscape.js · FastAPI · Python · Pandas · NetworkX · Docker
All wallets, labels, and cases are fictional; detected patterns are investigative leads, not attribution.
Live demo · Architecture · Tests
An end-to-end event platform connecting a React operations console to a FastAPI ingestion API, Redis Streams workers, and PostgreSQL analytics.
React · TypeScript · FastAPI · PostgreSQL · Redis Streams · Docker · Prometheus
Demonstrates atomic ingestion idempotency, durable deduplication, retries/DLQ, worker crash recovery, and frontend/API integration. Includes reproducible benchmark tooling without claiming unmeasured throughput.
Architecture · Benchmark method
An observability reference stack with service metrics, structured logs, alert routing, rule-based incident classification, and runbook-oriented response recommendations.
Python · FastAPI · Prometheus · VictoriaMetrics · Grafana · EFK · Alertmanager · Kubernetes · NVIDIA DCGM
Docker Compose exercises simulated inference telemetry locally. Kubernetes manifests provide a DCGM integration path for GPU-equipped clusters; this project does not claim production GPU-cluster operations.
A reusable real-time multiplayer systems prototype focused on server-authoritative state, ordered WebSocket events, fairness-sensitive randomization, layered settlement logic, React/PixiJS rendering, Redis reconnect sessions, PostgreSQL event history, and explainable integrity-risk controls.
TypeScript · Node.js · WebSocket · React · PixiJS · Redis · PostgreSQL · CSPRNG
A portfolio lab exploring the reusable architecture behind browser-based table and turn-based multiplayer systems.
An explainable access-monitoring pipeline based on behavioral baselines and rule-based risk scoring over synthetic access logs.
Python · Pandas · Detection Rules · Behavioral Analytics · Tests/CI
A synthetic fraud-risk engine combining customer baselines, transaction rules, velocity detection, and transparent scoring.
Python · Pandas · Fraud Analytics · Rule Engine · Tests/CI
- JavaScript / TypeScript / Python: primary development languages
- Frontend: React, Vue.js, HTML/CSS, browser/DOM programming, responsive UI
- Backend / API: Node.js, FastAPI, REST APIs, asynchronous processing
- Data: PostgreSQL, MySQL, Redis, SQL, Pandas, NetworkX
- Infrastructure: Docker, Docker Compose, Kubernetes foundations, Prometheus/Grafana
- Quality: pytest, GitHub Actions, reproducible local environments, synthetic test data
- Teaching: JavaScript application development and Python/data-analysis instruction through SSAFY/freelance teaching
- Built Python automation, crawling, data-processing, and AI-agent workflows in financial/forensic environments
- Developed a Python-based audio manipulation analysis workflow plus web-crawling and speech-to-text tooling
- Designed data-driven monitoring and detection logic with engineering/data teams
- Built and operated client-facing web/service workflows in a legal-tech environment
- Worked across product, engineering, data, security, compliance, and legal stakeholders to turn requirements into operational systems
I prefer projects that show more than isolated algorithms:
- a usable UI or API
- clear architecture and data flow
- reproducible setup
- tests and CI
- observable behavior
- explicit failure handling
- realistic production trade-offs
I use AI coding tools as an accelerator for implementation and review, while keeping architecture, debugging, validation, and final technical decisions human-owned.