I bridge the gap between advanced scientific modeling and practical systems engineering. I build computational physics simulations, design secure local-first productivity tools, and architect automated AI ingestion pipelines.

Interactive computational physics playground for simulating deterministic chaos and gravitational dynamics.

• Physics Engine: Models chaotic multi-body systems, Lagrangian mechanics (Double Pendulum), and stochastic polymers.
• Tech Stack: Python, NumPy, SciPy (DOP853 integrator), Matplotlib, Plotly, Streamlit.
• Engineering Highlights:
  • Resolved stroboscopic aliasing artifacts by decoupling dynamic temporal rendering steps from physics engine calculations.
  • Stabilized close-proximity gravitational encounters by implementing numerical softening factors ($\epsilon$) to prevent floating-point rounding errors and artificial energy creation.
• Demo: Live Streamlit App

Secure clinical natural language to SQL translator, observation pipeline, and RAG search dashboard.

• Tech Stack: Python (ChromaDB vector database, keyring), SQLite, Tailscale, Google Gemini API, Ollama (qwen2.5-coder:7b).
• Engineering Highlights:
  • Developed a vector-indexed synonym mapper inside ChromaDB to fuzzy-correct medical spelling anomalies and normalize observational test codes before query execution.
  • Designed an isolated WireGuard (Tailscale) network bridge to execute local inference calls securely without transferring clinical patient data to third-party public clouds.
  • Built a standalone credential management installer using the OS Keychain to prevent plaintext API token leaks at rest.

Automated AI-powered note classification, web scraping, and summarization pipeline for local knowledge graphs.

• Tech Stack: JavaScript (Obsidian API), Python (BeautifulSoup, re), Google Gemini API, Ollama (Local LLM), NotebookLM CLI.
• Engineering Highlights:
  • Developed a regex-based Unicode Math Bridge that dynamically translates LaTeX math syntax (e.g. $$x^T Ax$$ $\rightarrow$ xᵀ Ax) to native Unicode, preventing rendering crashes in Mermaid.js mind maps.
  • Designed self-healing state-machine rollbacks that reset failed source document uploads, ensuring zero pipeline deadlocks.
  • Implemented a dynamic NotebookLM Local Executable Resolver that parses Python's sys.executable path to execute virtual-environment binaries on Windows without global environment dependencies.
  • Engineered an Active Note Structuring & Auto-Renamer that scrapes target URLs, runs LLM-based categorization, formats Obsidian buttons, and dynamically sanitizes and renames note filenames for filesystem safety.
  • Integrated an HTTP Podcast Audio Server (podcast_server.py) that dynamically loads port settings from the plugin's configuration data and streams generated MP3s directly to media scrubbers.

Secure ingestion pipeline syncing Google Keep notes to a local relationship-driven Markdown vault.

• Tech Stack: Python (gkeepapi, keyring), JavaScript, Windows Credential Manager / macOS Keychain.
• Engineering Highlights:
  • Implemented a stateless OAuth token write-back mechanism that passes refreshed credentials strictly via memory environment variables and OS keychains, keeping active session keys entirely off the disk at rest.

Timeblocked calendar dashboard aggregating Google Calendar, Google Tasks, and Todoist with bidirectional sync.

• Tech Stack: Node.js, Todoist API, Google Tasks API, Tkinter, Gemini API.
• Engineering Highlights:
  • Handled Electron renderer thread blocks by wrapping API validation queries in an asynchronous Promise.race wrapper, enforcing graceful network timeouts.

Multimodal health dashboard, Bluetooth sync interface, and Git commit log aggregator plugin.

• Tech Stack: Python (Bleak BLE, OCR, Fitbit REST API), JavaScript (Obsidian API), Node.js child_process.
• Engineering Highlights:
  • Developed an interactive Bluetooth Pairing Wizard inside the Obsidian interface that manages local Bleak clients, unlocks secure Lorax protocol handshakes, and logs device telemetry (HRV, metrics).
  • Engineered a local-database-supported Food Registry & Ingestion Wizard that posts meals directly to the Google Health API (v4) and maintains a local dictionary of nutrition values.
  • Implemented clipboard screenshot OCR ingestion for logging workouts, focus logs, and daily scores using Google Gemini / local Ollama models.
  • Safely parallelizes native git log CLI fetches across multiple disconnected local repositories using asynchronous Promise.all executions without blocking the Obsidian Electron UI thread.

• Languages: Python, JavaScript (ES6+), Node.js, HTML5/CSS3, SQL, PowerShell
• Scientific Computing: NumPy, SciPy, Matplotlib, SymPy, Plotly, Streamlit
• AI & API Integration: Google Gemini API, Ollama, Google Health API v4, Todoist Sync, Fitbit REST API
• Security & Systems: Windows Credential Manager / macOS Keychain, Python keyring, Virtual Environments