Software Engineer / Juneau, Alaska

Charles McAndrews

A full-stack engineer leading the Azure migration of thirty-plus government systems while building accessible, AI-powered tools for education.

Five years in government technology. Senior contributions across .NET, Azure, React, and Swift. Known for shipping accessible, well-documented systems that teams can maintain after launch.

01 / Ask anything

Conversational resume.

For visitors who prefer to query rather than scroll. The assistant retrieves matching portfolio context, grounds the answer, and streams the response in real time.

Suggested prompts
charles.bot / ready

Hello - ask me anything about Charles' work, projects, or background.

02 / About

I build resilient, accessible software systems for organizations that cannot afford fragile infrastructure.

At the Alaska Department of Education, I lead modernization work across a legacy .NET portfolio: applications, pipelines, and production databases moving into Microsoft Azure with infrastructure managed through Terraform.

Prior to that role, I spent two years building the department's web platform: REST APIs, relational and document data layers, authentication, deployment automation, and accessible front ends.

Outside of work, I publish independent research and side projects: language-model fine-tuning, native iOS apps, and this portfolio's retrieval-augmented assistant.

I write production software with the assumption that it will be read, maintained, audited, and extended long after launch.

Based in
Juneau, Alaska
Position
SDE 3 - State of Alaska
Seeking
Senior engineering roles
Domain
Platform / AI / Govtech
03 / Selected Work

A representative sample of recent professional and independent work.

01
Alaska DEED Cloud Platform
Azure / Terraform / .NET 8
● Production
A cloud modernization program for Alaska education systems, moving legacy .NET applications, repositories, CI/CD pipelines, and databases from on-premise infrastructure to Azure.
ASP.NET CoreC#AzureTerraformCI/CD
2023-
→
02
Neuron Stream Lab
Transformer interpretability / WebGL
● Live
An interactive lab that runs a tiny transformer decoder in the browser and visualizes its activations, attention, and next-token distribution in real time — built to make a language model’s inner mechanics legible at a glance.
JavaScriptWebGLCanvas 2DTransformer
2026
→
03
AI Portfolio Assistant
Retrieval-augmented generation
● Live
The conversational interface on this site. It performs embedding retrieval over a curated professional knowledge base, grounds answers in matching context, and streams responses through the Vercel AI SDK.
Next.jsTypeScriptGPT-4o-miniEmbeddingsVercel AI SDK
2025
→
04
Personal Knowledge Model
Fine-tuning / LoRA / Apple Silicon
● Research
A LoRA fine-tuning system for adapting a 7B language model to a private personal corpus, with an Apple Silicon-optimized training pipeline and small adapter output.
PythonPyTorchMistral-7BLoRA / PEFTMPS
2025
→
05
MiClean
Swift / Vision / iOS
● App Store
A native iOS app that uses GPT-4 Vision to analyze a room photo and produce a personalized cleaning task list, with local task management built in SwiftUI.
SwiftSwiftUIGPT-4 VisionCore Data
2024
→
04 / Experience

Professional history.

Software Development Engineer 3

Jan 2023 - Present
State of Alaska - Dept. of Education & Early Development
  • -Leading the Azure migration of 30+ repositories, CI/CD pipelines, and databases from on-premise infrastructure.
  • -Modernizing legacy .NET Framework applications to .NET 8 with Terraform-managed cloud infrastructure.
  • -Building AI-powered educational tools with Azure OpenAI Services.
  • -Maintaining WCAG 2.0 accessibility compliance across public-facing and internal applications.

SDE 1-2 / Webmaster

Jan 2021 - Jan 2023
State of Alaska - Dept. of Education & Early Development
  • -Built and maintained full-stack applications with ASP.NET, C#, React, Angular, and Vue.
  • -Set up CI/CD pipelines and containerized deployments with Docker.
  • -Designed RESTful APIs and database solutions across SQL Server, PostgreSQL, and MongoDB.
Languages
C# / Python / TypeScript / Swift / SQL
Cloud / Infra
Azure / Terraform / Docker / CI/CD
AI / ML
Azure OpenAI / GPT-4o-mini / Whisper / RAG / Core ML
Front-end
React / Next.js / SwiftUI / Tailwind
Back-end
ASP.NET / Node / FastAPI / EF Core
Data
SQL Server / PostgreSQL / MongoDB / Redis
05 / Now

Active focus areas.

In progress

Hardening Azure migration patterns for legacy state education systems.

State of Alaska / Azure / Terraform
Researching

Local-first inference and the smallest viable model footprint for practical education tools.

Core ML / MLX / llama.cpp
Building

Small, useful AI interfaces that explain professional context without making recruiters dig.

RAG / Next.js / Vercel AI SDK
06 / Research

Published research.

arXiv:2604.21950April 2026

Feedback Over Form: Why Execution Feedback Matters More Than Pipeline Topology in 1-3B Code Generation

A study of small language model pipelines for code generation. Execution feedback improves results by more than four standard deviations on both benchmarks tested; a simple generate-execute-refine loop outperforms more complex pipeline topologies, and the refiner model’s quality matters more than the generator’s size.

cs.SEcs.AIcs.LGCode generationSmall LLMs
Read paper ↗