AI/LLM and geospatial/GPU software engineer. Rust and TypeScript. I build production LLM systems, GPU map rendering, and the cloud infrastructure that runs them. Lately: primary author of a safety-critical aviation flight-planning desktop app, lone architect of a real-time analytics platform. Remote, based in Waco, TX.
Not the Doom guy. Different guy.
- Production AI/LLM applications - schema-validated data extraction, LLM gateways, structured output, and the eval harnesses that keep them honest.
- High-performance maps and GPU/geospatial - deck.gl, luma.gl, MapLibre, WebGL, vector tiles and PMTiles.
- Aviation software - safety-critical flight planning and dispatch, on the desktop.
- Rust systems and cloud - Axum/Tokio services, type-safe Rust to TypeScript, serverless on AWS (Lambda, CDK), infrastructure-as-code.
AI / LLM
- llm-extract-evals - schema-validated LLM data extraction with a confidence-gated evaluation harness: accuracy scoring and offline replay.
- promptward - an LLM gateway that catches prompt injection and data exfiltration, validates structured output, and meters cost. The eval harness proves the detection rate.
Geospatial / GPU
- stormdeck - live weather on a deck.gl map, vector tiles served from AWS Lambda on the free tier.
- deck-wind-layer - a custom deck.gl v9 layer: GPU-advected wind particles over a moving map.
- glslint - a GLSL checker and language server that understands deck.gl/luma.gl shader modules (Rust).
Rust / systems
- typed-geojson - strongly-typed GeoJSON for Rust (
Feature<G, P>/FeatureCollection), specta-compatible. - specta - contributor to specta-rs, exporting Rust types to TypeScript.
- accept-payments - a payments and invoicing API in Rust/Axum on the AWS Lambda Rust runtime.
Rust, TypeScript, React, Next.js, Node, Tauri, Python | deck.gl, luma.gl, MapLibre, WebGL, D3 | Axum, Tokio, Snowflake, PostgreSQL | AWS (Lambda, CDK), Terraform, GitHub Actions | LLM/AI: extraction, evaluation, structured output