Walt Burge — AI Developer (Data Science & Learning Systems) · Oxford, MS
I’m a self-taught AI developer in Oxford, Mississippi, focused on data science and AI learning systems. I wrote my first line of code seven months ago; since then I’ve trained a custom LLM end to end — data curation, synthetic distillation, fine-tuning, and serverless deploy — built the tooling and agent systems around it, and shipped 11 production systems that put real AI in front of users. Open to AI / ML engineering roles.
Skills
- ML / AI — PyTorch, LLM fine-tuning (Llama 3.1), distillation, RAG, embeddings, LangChain, Ollama, MCP
- Data Science — dataset curation, synthetic data generation, evaluation, data pipelines
- ML Infra — RunPod, vast.ai (GPU), serverless inference, model ops
- Languages — Python, TypeScript, Java, Rust
- Backend & APIs — Spring Boot, Node.js, REST, WebSocket, PostgreSQL, SQLite
- Frontend — React, Next.js, React Native (Expo), Tailwind CSS
Contact: jamesburge.mcm@gmail.com · (662) 292-5533 · github.com/Aphrodine-wq
Selected Work
- FairTradeWorker — Two-sided construction marketplace connecting homeowners with vetted contractors. Next.js, Java/Spring Boot, React Native. QuickBooks-native payments and AI-powered estimation.
- MsHomePros — Contractor business platform with AI-powered line-item estimation, professional proposals, and job tracking. Live client: MHP Construction, Oxford, MS.
- ConstructionAI — Fine-tuned Llama 3.1 8B for construction cost estimation. 18,000+ training examples. Deployed on RunPod Serverless at ~$0.002 per estimate.
- AEON — A formal-verification and code-analysis platform: 73 engines (22 cybersecurity) that prove code correct and hunt vulnerabilities before it ships. Python.
- Engram — MIT-licensed, local-first screen memory: on-device OCR plus an MCP server, installable with one pip command. Python.
- Tessera — A markdown-native programming language for AI agents. Agents are written in .t.md files as substrate-tagged code fences, formally verified by AEON before they run. Python.
- W.A.L.T. — Distributed AI platform across three machines. 64 MCP tools for screen understanding, 73-engine formal verification, autonomous overnight execution.
Blog — The Build Log
-
Tessera: Why I'm Writing AI Agents in Markdown
— Building an AI agent shouldn't require five frameworks, three vendor SDKs, and a vector DB you babysit. It should require markdown and a compiler that takes the substrate boundaries seriously. The case for .t.md files as the source of truth for agents.
-
Substrates: Giving an Agent Named Modes of Thinking
— Tessera agents are built from typed code fences — logic, agent, memory, prompt, tool, neural. The compiler enforces the boundaries between them, so an agent's architecture becomes legible and auditable instead of a blob of Python.
-
Verifying AI Agents Before They Ever Run
— Tessera lowers each agent to a Substrate IR and runs AEON's 73 formal-verification engines against it before execution, catching capability leaks and PII egress at compile time. No other agent framework does formal verification before run.
-
Cognitive Traits: Channeling Maladaptive Patterns into Better Reasoning
— Tessera lets you install reasoning postures — productive doubt, cross-domain scanning, compulsive verification — as first-class, inspectable code in a tsr:traits block, instead of burying them in prompt strings.
- From Construction Sites to Codebases: 7 Months Self-Taught
- Fine-Tuning a Construction Estimation LLM from Scratch
- Designing a Three-Sided Construction Marketplace
Tech Stack
Next.js, TypeScript, Kotlin, Spring Boot, React Native, Python, Rust, Bevy, PostgreSQL, Tailwind CSS, PyTorch, Llama, RunPod
Contact
Email: jamesburge.mcm@gmail.com
GitHub: github.com/Aphrodine-wq