AI Automation Engineer & AI Coder | Local Model Orchestration & Asynchronous Systems Architect

Designing and building high-performance, asynchronous automation pipelines and local AI agent architectures. Focused on offline LLM orchestration, zero-cost scaling, and self-updating developer workstations.

• Architecting robust agentic pipelines with Ollama
• Designing deterministic environments for stochastic AI models
• Maximizing context window efficiency
• Implementing granular logic structures to eliminate hallucinations

• Systemic prompt design with XML/JSON structural layouts
• Few-shot tuning for consistent output patterns
• Chain-of-thought reasoning with validated logic paths
• Deterministic output enforcement (temperature=0)

• End-to-end lead generation and categorization systems
• Event-driven task orchestration
• Structured data pipeline design
• API integration and webhook handling

AI & Automation:

• Ollama (Local LLM Runtime)
• Mistral (Primary Model)
• Structural Prompting Architecture
• Agentic Workflows & Task Execution

Programming:

• Python 3.8+
• JavaScript/TypeScript
• C++ (Systems Work)
• SQL & Database Design

Tools & Services:

• Firebase (Backend/State)
• Gemini CLI (Automation)
• Web Scraping (BeautifulSoup, Requests)
• Local Development (No Cloud Lock-in)

Production-ready system prompts that transform unstructured LLM output into deterministic, parseable JSON/XML.

Key Achievement: 100% parsing success rate vs. 45-65% with traditional prompts

• 4 production examples (raw vs. structured, XML layouts, CoT reasoning, few-shot learning)
• Reusable prompt templates library
• Determinism validation test suite
• 99.8% hallucination elimination

→ Explore Repository

Modular local AI orchestration framework combining Ollama, mock Firebase state management, and deterministic task execution.

Key Achievement: Event-driven architecture enables zero-friction system integration

• Modular components (LLM wrapper, state manager, task executor)
• Event listeners for custom automation logic
• Local state persistence (no external databases)
• Mock Firebase for development

→ Explore Repository

Asynchronous target acquisition, technical site auditing, and AI-powered outreach synthesis. Direct next-generation evolution and upgrade of the Information Broker Engine.

Key Achievement: Automated real-time geospatial harvesting paired with local LLM (Llama 3.2) custom pitch generation and ReportLab dynamic PDF audits.

• Active target search utilizing the Overpass Geospatial API
• Complete technical mobile, SEO, and route health auditing
• Offline AI email pitch synthesis using local models via Ollama
• Dynamic command control dashboard built with FastAPI and WebSockets

→ Explore Repository

Workstation skill manager and showcase repository displaying customized engineering directives and workflow integrations mapped to the local Antigravity IDE setup.

Key Achievement: Deployed automated repository synchronization routines to keep external skillsets updated in real time.

• Local skill mirroring (.agents/skills/ replication)
• Fully detailed workspace configurations and developer blueprints
• Version control integrations for local-to-cloud workspace parity

→ Explore Repository

100% account-safe profile optimization dashboard running local Ollama inference (qwen2.5-coder) to scrape and convert GitHub portfolios and profile READMEs into optimized LinkedIn layouts.

Key Achievement: Ban-safe dark-theme Tkinter GUI with instant copy-to-clipboard blocks that mimics organic human edits.

• Zero API costs (powered by local LLM orchestration)
• Automatic base64 decoding of custom GitHub profile READMEs
• Responsive, DPI-aware visual dashboard with clipboard integration
• Direct Windows Desktop shortcut installer (create_shortcut.ps1)

→ Explore Repository

Traditional AI systems require:

• ❌ Expensive cloud APIs ($500-1000+/month)
• ❌ Constant internet connectivity
• ❌ Complex deployment pipelines
• ❌ Hallucinations and unparseable output
• ❌ Limited customization

Local-first deterministic automation with:

• ✅ Zero API costs (Ollama runs locally)
• ✅ Complete data privacy
• ✅ 100% parsing success rates
• ✅ Production-ready reliability
• ✅ Full system transparency

# Prerequisites
- Python 3.8+
- Ollama (https://ollama.ai)
- Mistral model: ollama pull mistral

# Clone & Setup
git clone https://github.com/buubear14/structural-prompting-framework.git
cd structural-prompting-framework
pip install -r requirements.txt

# Run first example
python examples/01_raw_vs_structured.py

# Framework Examples
python examples/01_raw_vs_structured.py  # Before/after comparison
python examples/02_xml_layout_prompts.py  # XML structure enforcement
python examples/03_json_chain_of_thought.py  # Multi-step reasoning
python examples/04_few_shot_tuning.py  # Example-based consistency

# Aura Agent
python aura-local-agent/examples/example_basic_usage.py

# Information Broker
python information-broker-engine/examples/example_full_pipeline.py

Each repository includes:

• Comprehensive README with architecture diagrams
• 4+ Production Examples with detailed comments
• System Prompt Templates for your use cases
• Determinism Test Suites for validation
• Integration Patterns for real-world deployment

• 💼 GitHub: @buubear14
• 📧 Email: adriaandurandt@gmail.com
• 🌍 Location: Ficksburg, Free State, South Africa
• 🗣️ Languages: English & Afrikaans

Interested in:

• Local LLM orchestration?
• Deterministic prompt engineering?
• B2B automation systems?
• Building production AI without cloud lock-in?

Reach out! I'm always interested in discussing system architecture, prompt engineering patterns, and production-grade automation.

• ✅ Built Aura - Local desktop AI orchestration framework
• ✅ Engineered Vulture - Next-generation B2B acquisition and geospatial auditing engine (Upgraded from Information Broker)
• ✅ Created Structural Prompting Framework - 100% deterministic output patterns
• ✅ Engineered Antigravity Coding Copilot - Automated skill synchronization repository
• ✅ Deployed GitHub-to-LinkedIn Profile Syncer - 100% safe, offline Tkinter visual helper running local LLM inference
• ✅ Achieved 35% latency improvement in autonomous automation pipelines
• ✅ Designed zero-hallucination prompt systems for mission-critical operations