Systems Engineer | AI Strategist | Infrastructure Architect
Standardizing the telemetry and extraction layers of agentic AI, with resilient defaults and strict contracts.
I build systems where the weakest link is assumed to be:
- the user
- and now local models that are not very smart
So the defaults must be safe, the interface must be strict, and the system must keep moving when parts fail.
High-concurrency local compute for orchestration, deep-research automation, and local inference.
- Dell PowerEdge T630
- 80-core CPU, 160GB Registered DDR4
- Dual Tesla P40 (48GB VRAM total)
- Hybrid Windows + Ubuntu
- Ollama-first workflow, async-first Python stacks
Resilient, fault-tolerant tooling for messy environments and unreliable inputs.
- Agentic telemetry: async-first observability that does not block the agent loop, with reasoning trace capture (example:
<think>) - Structured extraction: strict JSON Envelopes to reduce context pollution and improve tool-call reliability
- Resilience engineering: retries, backoff, validation, partial-result preservation, and strategy failover (including headless to visible)
Curated entry points. Repos contain the implementation details.
- Web Scraper Toolkit (MCP-native extraction): https://github.com/imyourboyroy/WebScraperToolkit
- LLM Telemetry Toolkit (async-first observability): https://github.com/imyourboyroy/llm-telemetry-toolkit
- Email Toolkit (parsing + diagnostics utilities): https://github.com/imyourboyroy/emailtoolkit
- ReqSync (repeatable Python environments): https://github.com/imyourboyroy/reqsync
I treat software, infrastructure, and the human psyche as the same class of problem: constraints, feedback loops, and measurable outcomes.
- Logic-driven strategy and psycho-relational analysis
- Active LDS, faith framed as practical and measurable benefits to the psyche
- Biohacking without magical thinking, iterate with data