Zain Dana Harper - Software Engineer

Current state

Small proof-surface review.

Public release review

Claim-to-evidence mapping for public surfaces.

A scoped review for one repo, repo cluster, AI workflow, or launch surface. I map what the surface claims, what evidence is visible, what is missing, and what should change before a customer, investor, reviewer, or governance handoff sees it.

7-day reviewpublic or supplied surfacesmaturity notes

Not exploit testing. Not certification. No credentials. No production access.

Request review Sample report

Toolchain map

Surface -> provenance -> proof -> report -> witness.

The stack is intentionally narrow. Each tool occupies one point in the release pipeline: inspect the public face, validate provenance records, index evidence, assemble a report, and keep EMET as a witness rather than an authority.

inspectvalidateindexreportwitness

public-surface-sweeper model-provenance-validator repo-proof-index warden-reporting EMET

public-surface-sweeper

Release hygiene point

Checks what a repo says before the internet, customers, or reviewers see it: public claims, release hygiene, secret-shaped values, docs gaps, and presentation risk.

public repo surface · pre-release gate · findings and fixes

GitHub

model-provenance-validator

Provenance envelope point

Validates compact model provenance records so model and workflow claims stay attached to inspectable evidence.

model record · release review · pass/fail with gap notes

GitHub

repo-proof-index

Evidence assembly point

Turns scattered receipts, schemas, tests, and release artifacts into one proof index that a human can inspect quickly.

proof artifacts · audit packet · markdown and JSON

GitHub

warden-reporting

Human handoff point

Structured evidence chains, finding correlation, and risk-scored reports from sanitized WARDEN tooling.

evidence · findings · reviewer handoff

GitHub

EMET

Witness point

External source/view consistency checks. EMET says MATCH, DRIFT, or UNVERIFIABLE; it does not say TRUSTED.

source/view pair · model-facing boundary · witness receipt

GitHub

Selected work

Systems, surfaces, and experiments with labels.

Real-time graphics · HLSL / Python

ELDER.ENB

A real-time shader post-processing framework for Skyrim: 155+ shader overlays, a color-graded preset system, and Python tooling. It was my first public graphics release, and I learned most of it by getting it wrong first.

887K+ downloads149K unique2,378 endorsements

HLSL · color grading · shader pipelines · Python automation · April 2024

Nexus Mods Demo video

Compilers · Rust · public, MIT

QuantaLang

An effects-oriented compiler built around quantac. The public state is labeled plainly: C is the verified execution path, HLSL/GLSL shader output works, and the other targets are experimental research surfaces.

656 testsC backend verifiedHLSL / GLSL output

effects language · C99 · semantic receipts · crates.io v1.0.0

Landing page GitHub VS Code Marketplace

Harper Advocates & Harper Compliance

Live frontend surfaces · private GitHub repos

Two live service sites I designed and shipped: Harper Advocates for consumer-rights letters and Harper Compliance for compliance-documentation kits. The repos are private; the public surfaces show responsive UI, custom cursors, pricing/category cards, conversion paths, and browser-side document workflows.

live websites · custom cursors · responsive UI · private GitHub repos

Advocates
Compliance

RAW (Rendering Advancement Workshop)

Real-time graphics · D3D11 / C++23 · private prototype

A standalone Direct3D 11 post-processing injector: physically-based tonemapping, TAA, SSR, volumetrics, bokeh DoF, and ImGui controls in a self-contained DLL.

18+ effects · 86 HLSL shaders · ~45K LOC · ACES / AgX

private

Calibrate Pro

Color science · Python · MIT

Sensorless display calibration research: ICC v4.4 profile generation, 3D LUT output, and DDC/CI hardware control experiments.

Delta E 0.65 target · ICC v4.4 · 3D LUT · DDC/CI

GitHub

Quanta Universe

Systems · QuantaLang · alpha surface

A broader demonstration ecosystem for Quanta research across rendering, color, finance, systems, and tooling modules. It is a proof surface and alpha corpus, not a whole-ecosystem production build.

23 modules · OS kernel experiments · 64+ utility ports · self-hosting research

GitHub

Quanta Color

Color science · Python · MIT

A color-science library: 15+ color spaces, 12 tone mappers, CIEDE2000/Oklab difference metrics, CAT16 chromatic adaptation, color-vision-deficiency simulation, and ICC profile generation.

CIECAM · Oklab · spectral workflows

library

Aurora

Tooling · Lua / TypeScript / DSL

A combat-scripting SDK with a custom indentation-sensitive DSL compiler, 67 SDK modules, production rotations, and a simulation test harness.

4-stage DSL compiler · TSTL · 39 specs / 13 classes · v1.0.0

SDK

Multi-Agent LLM Orchestration

AI systems · Python · local/private platform

A director-pattern platform that classifies tasks and routes them to specialist agents, with meta-agents for decomposition, coordination, validation, and synthesis over a multi-provider model abstraction.

director pattern · specialist agents · MCP servers · multi-provider routing

platform

EMET

AI safety · Python / Rust / Node.js · MPL-2.0

A re-derivability-based integrity verifier for AI oversight: it checks that bytes reaching a model match their source and flags when a presented view differs from what should be read. Verdicts are MATCH, DRIFT, or UNVERIFIABLE.

3 implementations · conformance suite in CI · draft spec · STRIDE threat model

GitHub

AI-safety research

Defensive work on model-facing proof and context integrity.

I write defensive notes on prompt and context-integrity failure modes in LLM coding assistants, including authorization-laundering / vocabulary-calibration as a safety-assessment pattern. The public primitive is EMET, a witness tool for source/view consistency. Operational exploit detail stays out of the public page by design.

Approach

How the work gets made.

My path is nontraditional: commercial arborist, Xbox networking support, then a sustained self-directed engineering sprint across compilers, graphics, color science, frontend sites, and AI tooling.

I use agentic coding tools heavily, but not as a substitute for ownership. I set the scope, review diffs, run tests, keep public/private boundaries clear, and label maturity honestly.

Small claims. Visible receipts. Honest maturity labels.

Experience

Where it has taken me.

2023 - Present
Remote

Independent Software Engineer & AI-Safety Researcher

Compilers, real-time graphics, color science, multi-agent LLM systems, and two live frontend product surfaces, shipped across GitHub Pages, crates.io, the VS Code Marketplace, and Nexus Mods.

2015 - Present

Commercial Arborist · Legendary Tree (family business)

Operations, site assessments, safety procedures, and written client proposals across 11 years.

2013 - 2014
Redmond, WA

Technical Networking Support, Xbox Division · contract via Convergys/Stream

Diagnosed TCP/IP, DNS, NAT, and firewall issues for Xbox Live networking on a vendor contract; contributed to internal knowledge-base documentation.

Toolkit

What I work in.

PythonRustC++TypeScriptLuaHLSLC#BashCompiler designDirect3D 11Real-time renderingGPU programmingColor scienceStatic frontend sitesAccessibility passesDesktop GUILLM orchestrationMCP serversProvenance & integrityCI/CD