Data Engineering & Apps
End-to-end data pipelines, analytics infrastructure, and live products I've built and shipped.
Gaming Analytics Pipeline — dbt + BigQuery + Airflow
End-to-end ELT pipeline built on real GA4 and AppsFlyer data from live mobile game titles. Covers the full modern data stack from raw event ingestion to business-ready marts.
Features: Incremental models · Snapshots (SCD Type 2) · Custom generic tests · dbt docs & lineage · Airflow DAG orchestration · Data quality monitoring
AI E-Commerce Chatbot — LangChain + GPT-4o + FAISS
LLM-powered conversational chatbot for e-commerce product discovery. Built with RAG architecture using FAISS vector store for semantic search over product catalog data.
Features: Semantic product search · Conversational memory · Streamlit interface · Deployed end-to-end
Game Analytics Infrastructure — Newga Network
Designed and maintained the full analytics data infrastructure for mobile games serving 10M+ global users. Owned everything from event instrumentation to executive dashboards.
Impact: D7 retention +15% · ARPU +10-20% · Analytics turnaround -30% · Supported $15M+ acquisition
I build and ship mobile games and AI apps independently — and own the full data infrastructure for each title, from GA4 instrumentation to dbt-modeled analytics marts.
SnaPet
A consumer social app where pets come alive through AI. LLM + computer vision give each pet a unique personality that responds to photos and chats.
Poker Mahjong
A hybrid casual puzzle fusing poker and mahjong tile-matching. Full product ownership from design through live operations and analytics.
21 Saga
Number-matching puzzle combining blackjack mechanics with satisfying tile-clear gameplay. Owned full product lifecycle from concept through analytics.
Tile Voyage
Zen tile-matching with island-building progression. Designed engagement loops and retention mechanics, tracked via full analytics pipeline.
Mahjong Monopoly
Strategic mahjong fused with board game territory mechanics. Android live; iOS under App Store review.
Cinema Analytics — Multi-Stakeholder Perspective
Analyzed box office trends through three business lenses. Produced a film where each character visualizes the same dataset differently.
Stock Price Forecasting & Investment Strategy
Time series modeling of AAPL and HON using ARIMA, regression, and moving averages. Compared simulated trading strategies with comprehensive performance analysis.
↗ GitHubSports Betting Strategy — Statistical Modeling
Monte Carlo simulation and chi-square testing to find optimal betting strategies across best-of-3/5/7 series formats.
↗ GitHubCoffee Price Analysis — Statistical Forecasting
EDA, ANOVA, and Lasso regression on coffee bean prices. Includes STL decomposition and MAPE-evaluated forecasting models.
↗ GitHubModel Validation — K-Fold Cross-Validation
Compared Logistic Regression, Random Forest, and SVM using K-Fold and Repeated K-Fold to investigate performance stability.
↗ GitHubUK Bicycle Accident Analysis Dashboard
Interactive Looker Studio dashboard analyzing 40 years of UK bicycle accident data to surface policy-relevant insights.
↗ GitHub