CS/Software Engineering graduate (January 2026), full-stack and AI/ML engineer. Open to full-time opportunities.
I build at the intersection of full-stack development and applied ML — agents with tool use, prediction pipelines, desktop software, deployed demos. My work tends toward projects that ship and can be inspected end to end. Based in Minnesota.
soundEQ — System-wide parametric equalizer for Windows
Tauri v2 desktop app that captures all Windows audio via WASAPI loopback, applies real-time DSP, and routes the processed audio to your speakers. Supports up to 16 parametric bands per profile across 7 filter types, per-app profile routing that auto-switches on window focus, a live FFT spectrum analyzer at ~20 Hz, headphone crossfeed, and 50-step undo/redo. Shipped as a signed v1.0.0 installer.
Stack: Rust, React, TypeScript, Tauri v2, WASAPI, Tailwind.
Why it matters: Real audio systems work — biquad filter chains, low-latency loopback I/O, device hot-swap recovery — wrapped in a polished cross-process desktop app. Demonstrates ability to own a project from DSP core to UX detail.
GardenMind — AI gardening assistant with tool use
Streamlit chat agent built on Groq's llama-3.1-8b-instant that chains six tools — live weather,
5-day forecast, USDA frost dates, a 200-plant care library, and per-user garden profiles — to
answer questions like "Is it safe to plant tomatoes in Denver this weekend?" Includes a 25-case
evaluation suite (currently 80% pass rate) and a UI that shows every tool call in real time.
Stack: Python, Groq API, OpenWeatherMap, Streamlit.
Why it matters: Modern AI engineering, end-to-end. Tool-use agent design, custom evals as a first-class deliverable, transparent execution, and a deployed live app at gardenmind.streamlit.app.
weatherEdgeAI — AI backend for weather prediction market analysis and paper trading
FastAPI backend that discovers weather-related prediction markets via Polymarket's Gamma API, parses questions into structured weather targets, fetches Open-Meteo forecasts, and estimates resolution probabilities with a logistic regression model. Expected value is evaluated against live market prices to generate paper trade signals; backtesting reports include Brier score, log loss, calibration buckets, paper ROI, and max drawdown. A React dashboard surfaces the full pipeline — market discovery through paper trade status — with a public dry-run mode so reviewers can inspect live signal validation without creating trades.
Stack: Python, FastAPI, PostgreSQL, SQLAlchemy, Open-Meteo, React, TypeScript, Docker.
Why it matters: Shows the full stack of applied probability engineering — structured data pipelines, a trained prediction model, EV-based strategy logic, and rigorous backtesting — in a production-shaped backend with explicit safety boundaries between paper and live execution.
Blackjack AI Decision Assistant — Monte Carlo + supervised ML decision assistant
Browser-based Blackjack assistant that recommends hit, stand, or double down with explainable odds. The first layer is a Monte Carlo simulator computing expected value from the visible game state; the second is a supervised-learning workflow that generates a 10,000-row labeled dataset, engineers features, and trains decision-tree and random-forest classifiers (random forest reaches ~91% test accuracy). Deployed at dcass5212.github.io/blackjack-ai-decision-assistant.
Stack: JavaScript, HTML, Python, scikit-learn.
Why it matters: Clean end-to-end ML — dataset generation, feature engineering, model training, evaluation, inference — paired with an interactive demo that makes the model's reasoning visible.
Languages: Rust, Python, JavaScript, TypeScript
Frameworks and libraries: FastAPI, React, Tauri, Streamlit, scikit-learn, SQLAlchemy, Pandas
Tools and infrastructure: PostgreSQL, Docker, Vite, GitHub Actions, Alembic
AI/ML: Groq API, tool-use agents, logistic regression, random forests, Monte Carlo simulation