Rob Simpson pertrai1

Rob Simpson

Software Engineer with 15+ years experience.

I thoroughly enjoy Meta-Learning - understanding the mechanics of how we acquire and apply new knowledge.

My current focus is "AI Engineering" - understanding the low level design of RAG, embeddings, and model orchestration.

PLANTS NLQI - A Natural Language Query Interface for the USDA botanical database. Solving for accuracy in specialized scientific domains using RAG.
OpenSpec Architecture - Experimenting with declarative, YAML-based AI agent definitions to ensure developer-led control over agentic workflows.
Model Translation - Learning to port Python ML models to browser-ready TypeScript/TensorFlow.js to enable high-performance, client-side intelligence.
508 Accessibility Tools - Prototyping applications that help developers automate and verify Section 508/WCAG compliance in modern SPAs.

AI Projects (ai-projects repository): Developed and experimented with various AI concepts including:
- FieldGuide Assistant: An AI-powered assistant for botanical information.
- Agent Orchestration: Exploring multi-agent systems and their coordination.
- Extensive documentation and research on Agents, RAG, LLMs, N-Shot Learning, and related AI topics.
- Research Paper Implementations: Developed small-scale projects based on key research papers in the AI field to validate concepts and explore implementation details.
Contributions & Challenges (coding-challenges repository):
- Implemented a 'GFE Custom EventEmitter'.
- Solved numerous Blind 75 and other LeetCode challenges, focusing on data structures and algorithms.

A lab for production-ready AI implementations.

FieldGuide Assistant: Multi-document RAG system focused on context-aware retrieval.
Agent Orchestration: Building frameworks that keep the human "in the loop," prioritizing developer agency over autonomous agent control.

A laboratory for problem-solving and meta-learning.

Learning How to Learn: Using challenges as a medium to refine Active Recall and Spaced Repetition techniques in a technical context.
First Principles: Breaking down complex problems into primitive patterns (Graph theory, Dynamic Programming, Sliding Windows) to build a reusable mental library for system design.
Process over Product: Each solution is an exercise in documenting the "why"—translating abstract requirements into clean, performant TypeScript/Python code.