# Arthur Perng

*Resume — software track.*

Software developer — ML, quantitative finance & full-stack

arthur@perng.com · [arthur.perng.com](https://arthur.perng.com) · [linkedin.com/in/arthurperng](https://www.linkedin.com/in/arthurperng/) · [github.com/Happygator](https://github.com/Happygator) · San Jose, CA (open to relocating anywhere in the US)

## Education

**University of Maryland** — College Park, MD — *Expected Jun 2027*
B.S. in Computer Science and Mathematics, Minor in Computational Finance
GPA: 3.94 / 4.00 · Dean's List

## Technical Skills

- **Languages:** Python, Java, C++, OCaml, SQL, TypeScript, C, R, Matlab
- **Libraries:** pandas, NumPy, Flask, PyTorch, matplotlib, FastAPI, Flutter
- **Developer Tools:** Microsoft Azure, Google Cloud, Firestore, Git, MongoDB, Claude Code, OpenAI Codex, Antigravity
- **Awards:** USA Coding Olympiad Gold, Top Reverse Engineering Competitor of US Cyber Open, AIME Qualifier
- **Topics:** LLMs, AI Agents, ML Systems, Data Pipelines, Backend/API Design, Data Visualization, Numerical Methods

## Experience

### Capital One — Machine Learning Intern
*McLean, VA · Jun 2025 – Aug 2025*

- Built an LLM routing system to direct 50K+ daily queries to optimal models.
- Designed the routing and ETL pipeline to make decisions in 200ms with 88% accuracy.
- Optimized 4+ endpoints, reducing infrastructure costs by $150K+ annually.
- Collaborated day-to-day with a fellow intern, a senior engineer, and a mentor to design, build, and ship the system.

### NASA Goddard Space Flight Center — Software Engineering Intern
*Greenbelt, MD · Jun 2024 – Aug 2024*

- Built a Python library and Jupyter notebook tutorial for NASA's TOLNet API, driving adoption to 1K+ daily requests.
- Organized API data using pandas and matplotlib to create curtain plots for ozone-layer visualization.
- Partnered with a mentor to scope and build the library and its documentation.

## Research

### UMD FIRE Research — *R, terra, tidyverse*
*Sep 2023 – May 2024*

- Evaluated how expansion of the Maryland Silver Line affected home rental prices using historical data.
- Wrote R code using terra and tidyverse to build linear regressions, with plots in matplotlib.

## Projects

### Google AI Agent Capstone Project — *Python, Google ADK, FastAPI, Flutter*
*Jun 2026 – Ongoing · [GitHub](https://github.com/Happygator/home-rescue-agent) · [YouTube](https://youtu.be/qbdIQUzFJYY)*

- Built a concierge AI agent using Google ADK and Gemini to troubleshoot appliances, an MCP-based tool server for OEM data lookups, and a rules-based escalation engine to ensure safety-critical guidance.
- Deployed a production FastAPI backend on Google Cloud Run with Firestore and Cloud Storage, including a PDF ingestion pipeline for large manuals.
- Developed a cross-platform Flutter client (Android, iOS, Windows) with vision recognition and chat UI.

### Polymarket Calibration Research — *Python, pandas, SQLite, DuckDB*
*Jun 2026 – Ongoing · [GitHub](https://github.com/Happygator/polymarket)*

- Built a resumable Python crawl/ETL pipeline to collect 1.34M Polymarket markets and 9.4M historical price snapshots into an indexed SQLite database queried with DuckDB over 5.9M rows.
- Computed calibration metrics including bucketed realized-vs-implied probability, Wilson confidence intervals, Brier score, log-loss, and Expected Calibration Error.

### Wynncraft Market Intelligence Dashboard — *TypeScript, React, SQLite*
*Jun 2025 · [GitHub](https://github.com/Happygator/Wynncraft-Market-Intelligence)*

- Built a Python ingestion pipeline to collect Wynncraft trade listings and store historical market snapshots in SQLite.
- Computed market metrics (VWAP, slippage, price impact, liquidity, volatility) and served them via a FastAPI backend to a React dashboard.

### Grocery Expense Analyzer — *Python, Gemini Vision, SQLite*
*Jan 2025 · [GitHub](https://github.com/Happygator/Receipt-viewer)*

- Built a Discord bot using Gemini vision + OCR to extract items and prices from grocery receipts, storing them in SQLite for future analysis.