# Parker Patton

574-371-6442 | jppatton@crimson.ua.edu | https://www.linkedin.com/in/jparkerpatton | https://github.com/catears124 | https://parker.ml

Canonical PDF: https://parker.ml/Parker_Patton-Resume.pdf

## Education

**The University of Alabama** — B.S. Mathematics, minor in Computer Science, August 2025–present. Expected graduation: May 2028. Tuscaloosa, Alabama.

Fall 2026 coursework: Data Structures & Algorithms, Software Engineering, Calculus III, Differential Equations.

## Experience

### AI Safety Lab — The University of Alabama

September 2026–present · Tuscaloosa, Alabama

- Developed a low-dimensional latent-space attack using video-to-video models for downstream classifiers; ran a 24-shard, 2,400-clip GPU campaign and achieved 94.5% attack success at 27–30 dB PSNR.
- Mapped decoder–classifier geometry with Jacobian and gradient projections; found 47.3× above-chance task-gradient concentration while 90.7% attack success persisted outside dominant decoder modes across 300 Kinetics-400 clips.

### LLM Interpretability Research — Algoverse AI Research

May–September 2026 · Remote

- Designed a content-matched 2×2 factorial experiment with over 11,520 matched prompts and three Llama checkpoints; extracted residual-stream representations at each transformer block and fit held-family-out linear probes.
- Identified a prompt-format confound: naive probes achieved 0.600–0.682 AUC, while format-controlled probes reached 0.989–0.998 AUC and generalized across unseen formats at 0.940–0.985 AUC.

### Financial ML Research — The University of Alabama at Birmingham

June 2024–May 2025 · Birmingham, Alabama

- Built a diffusion-empirical model for next-day return distributions across five assets and 45 walk-forward folds.
- Raised volatility R² from 0.179 to 0.197 and reduced CRPS by 0.30% versus EWMA across 14,175 out-of-sample forecasts.
- Presented at the ASFA Symposium; awarded Honorable Mention in CARSEF Math & CS.

## Projects

### tablet.ears.cat

Python, TypeScript, C, ARM firmware · September 2026–present · https://tablet.ears.cat · https://github.com/catears124/tablet.ears.cat

- Developed custom firmware for the Gaomon S620 graphics tablet and deployed a web installer to over 7,000 users.
- Reverse-engineered OEM firmware, increased polling from 266 Hz to 1,000 Hz, and eliminated over 15 ms of input latency.

### NullMap (HackMIT 2026)

Python, FastAPI, Elasticsearch, React · September 2026 · https://plume.hackmit.org/project/cuqqq-zpzje-qzyls-tmbjd

- Built a resumable ingestion and embedding pipeline indexing over 2.1 million OpenAlex and ClinicalTrials.gov studies.
- Combined BM25 and kNN retrieval to optimize LLM context, reaching 19 seconds and $0.021 per query with an 88% cost reduction.
- Deployed in 24 hours and won second place in Elastic's “Find the Signal” sponsor challenge at HackMIT 2026.

### osu!rankguess

Python, PyTorch, FastAPI, ONNX, PostgreSQL · March–August 2026 · https://github.com/catears124/osu-rankguess

- Built and published osu!3k, a multimodal dataset of 3,000 gameplay videos, replay files, and player metadata.
- Trained a CatBoost/computer-vision ensemble reaching 0.846 R² and 1.84% MAE when predicting percentile from a single sample.
- Deployed a FastAPI/ONNX platform on Vercel with replay parsing, CPU inference, and PostgreSQL storage.

### Independent Smart-Contract Security Research

Solidity, Foundry, EIP-712, HyperEVM · June 2026

- Responsibly disclosed two settlement vulnerabilities in live options contracts by reverse-engineering EIP-712 signatures, packed calldata, and balance-debit paths.
- Found 38 post-expiry settlements across 800 live transactions and reproduced both flaws on a mainnet fork.

## Technical Skills

- Languages: Python, C/C++, TypeScript, SQL, Solidity
- ML & Data: PyTorch, scikit-learn, NumPy, pandas, CatBoost, ONNX
- Backend & Web: FastAPI, PostgreSQL, Elasticsearch, Next.js, React
- Tools & Platforms: Git, Linux, Docker, Slurm, Vercel, GCP, AWS, GitHub Actions