Projects

Systems I've designed.

Open-source frameworks, tools, and architectural experiments. Each one addresses a specific design problem — and documents the trade-offs that shaped the solution.

Featured Project

POLQA

LLM Output Governance & Bias Evaluation Framework

A modular CLI framework for governing LLM output quality at scale. The core architectural decision was provider-agnostic design with a pluggable evaluation layer — chosen over a tightly-coupled solution because LLM ecosystems fragment rapidly and auditability requires reproducibility across model versions. Supports HuggingFace, Ollama, OpenAI, Claude, and custom endpoints.

Outcome

Enforces reproducibility, cross-model comparability, and auditability — governance properties absent from most academic and commercial evaluation tooling.

Python LLM Governance Adversarial Testing HuggingFace OpenAI

Automation Learn

Scalable QA Capability — Decoupling Knowledge from Individuals

A structured curriculum designed around a specific architectural principle: quality capability should be institutionalized, not person-dependent. The decision to build opinionated learning paths over encyclopedic references was deliberate — transferable judgment scales; tool familiarity does not. Covers Selenium, Cypress, API testing, CI/CD, and test architecture.

Outcome

Reduces knowledge-transfer and ramp-up time for engineers entering automation roles.

Selenium Cypress Robot Framework CI/CD GitHub Actions

SynLeo

Generative AI Architecture for Creative Systems

An architectural exploration of generative AI applied to music production. The key design question was whether AI functions better as a creative constraint or a generative layer — and what the system boundaries look like when the output is subjective. Covers voice modeling, lyric generation, and the governance challenges unique to AI-generated creative content.

Outcome

A working model for how generative AI can be integrated into creative workflows without displacing authorship.

Python Generative AI System Design Creative Tech

SPA — System for Thought & Action

Information Architecture for High-Complexity Knowledge Work

A personal operating system that synthesizes PARA, GTD, and SCRUM into a single coherent information architecture. The design trade-off was consistency over flexibility — a stricter schema reduces cognitive load at the cost of customization. Built for technical professionals managing concurrent projects, continuous learning, and long-horizon career decisions.

Outcome

Demonstrates that information architecture principles — hierarchy, separation of concerns, clear interfaces — apply equally to knowledge management systems.

Obsidian Information Architecture PARA GTD Systems Design

Resume Builder

Structured Data → Professional Document Pipeline

Enforces a clean separation between content and presentation — an architectural boundary most resume tools collapse, creating lock-in and brittle outputs. YAML as the single source of truth; PDF as a deterministic, reproducible artifact. Optimized for ATS parsing without sacrificing human readability, because both constraints must be satisfied simultaneously.

Outcome

Produces documents that parse cleanly through major ATS systems and communicate clearly to human reviewers.

Python PDF Generation CLI YAML

README Wizard

Documentation as a First-Class Engineering Artifact

Treats README generation as a structured pipeline problem rather than a writing task. The decision to constrain output format was intentional — open-ended documentation tools produce inconsistent results because they offer too many degrees of freedom. A minimal schema forces the four questions every README must answer: what it does, how to install it, how to use it, and how to contribute.

Outcome

Reduces time to first usable documentation from hours to minutes, with consistent structure across projects.

Python CLI Markdown Documentation Design

All projects are open-source. Issues, forks, and contributions welcome.

View on GitHub →