Applied AI Engineer | Agentic Systems Builder | Staff Full Stack Engineer
I turn messy real-world workflows into practical agentic systems, retrieval-backed assistants, and operator-facing tools that people actually use. 30 years of full-stack engineering depth — now focused on shipping dependable AI systems.
I design and ship multi-agent systems that run autonomously on daily schedules, scrape live data, make structured decisions, and surface results through dashboards and notifications — with humans in the loop where it matters.
┌──────────────────────────────────────────────────────────────────────────┐
│ OPENCLAW · multi-agent platform │
│ │
│ ┌───────────────────────────────────────────┐ │
│ │ ATLAS · orchestrator & cron scheduler │ │
│ └─────────────────────┬─────────────────────┘ │
│ │ │
│ ┌────────────────────────┼────────────────────────┐ │
│ │ │ │ │
│ ┌────────────────┐ ┌────────────────┐ ┌────────────────┐ │
│ │ Family Guy │ │ Lyra │ │ Trader │ + more │
│ │ household ops │ │ marketing │ │ crypto swing │ │
│ │ calendar RAG │ │ outreach │ │ signals │ │
│ │ daily briefs │ └───────┬────────┘ └───────┬────────┘ │
│ │ ┌────────────┐ │ │ │ │
│ │ │ Schooly │ │ │ │ │
│ │ │ grades │ │ ┌───────────────┐ ┌───────────────┐ │
│ │ └────────────┘ │ │ Bittensor │ │ Sales Mgr │ │
│ └────────┬───────┘ │ TAO staking │ │ workshop ops │ │
│ │ │ subnet alpha │ │ Wix API + │ │
│ │ │ research rig │ │ AgentMail │ │
│ │ └───────┬───────┘ └───────┬───────┘ │
│ │ │ │ │
│ structured deterministic API-driven │
│ file reports proposal engine email routing │
└──────────────────────────────────────────────────────────────────────────┘
| Skill | How I Apply It |
|---|---|
| Multi-Agent Decomposition | Each agent has one job, its own context window, bounded tools, and well-defined interfaces |
| Context Architecture | RAG pipelines, structured workspace files, and retrieval patterns — not stuffing everything into prompts |
| Specification Precision | Step-by-step heartbeat contracts that eliminate model drift in autonomous daily runs |
| Deterministic Tool Design | Risk gates, data normalization, and formatting run as code — LLMs only enter where reasoning is needed |
| Human-in-the-Loop Automation | Approval checkpoints, review queues, and graduated autonomy based on the cost of getting it wrong |
| Failure Pattern Recognition | Tracing through session logs to diagnose silent failures like models emitting wrong field names |
| Cost & Token Economics | Local models for cheap classification, cloud reasoning reserved for high-stakes decisions |
AI/Agents: OpenClaw · MCP · RAG · Multi-Agent Orchestration · LLM APIs · Local Models (LM Studio) · Prompt & Tool Schema Design · Browser Automation · Claude Code · Codex · Hermes Agent · Bittensor/TAO
Full Stack: Python · .NET/C# · TypeScript · Angular · React · Next.js · REST APIs · GraphQL
Infrastructure: SQL Server · Postgres · MySQL · AWS · Azure · Terraform · Kubernetes
Robotics: Arduino · PlatformIO · ESP32 · Raspberry Pi
📐 Agentic Systems Architecture — Detailed workflow diagrams, design decisions, and architecture docs for the multi-agent systems I've built. No source code — just the thinking behind the systems.
🌐 Portfolio · 📄 LinkedIn · ✉️ carpenter.jason.m@gmail.com