now — logan

current focus

Designing agentic systems for enterprise customers — multi-agent workflows that survive contact with production constraints.

Pre-sales architecture: discovery workshops, sandbox labs, executable prototypes for Director–CIO stakeholders.
Banking and financial-services engagements where agentic patterns replace brittle automation.
Chair of an internal Invention Discovery Team — peer review for novelty, feasibility, commercial fit.

shipping next

An agentic contract-analysis system for a major banking client, plus a reusable pattern library for agentic workflow augmentation across enterprise environments.

Production rollout milestones for the contract-analysis pilot.
Reference architectures for agent-to-tool and agent-to-agent handoffs.
Next invention-disclosure batch — Q3 2026.

growing into

Deeper depth on agentic evaluation — eval harnesses, governed-agent patterns, and the operational discipline that turns demos into systems customers can run on a Tuesday.

Comparative depth across agentic frameworks — LangGraph, CrewAI, AutoGen, and the broader orchestration landscape.
Tool-use, planning, and memory patterns across cross-system flows.
Multi-modal retrieval patterns for enterprise document corpora.

reading & studying

Papers on agentic evaluation, multi-agent orchestration, and the rapidly-moving agentic-framework landscape. Also enterprise-AI regulatory drafts as they land — the rules around governance are moving as fast as the models.

side experiments

A steady cadence of after-hours builds — mostly MCP tooling and coding-agent tuning. The goal is craft: get faster at the parts of the loop that are mine to control.

Custom MCP servers — wiring local context, internal docs, and project state into the agents I use daily.
Coding-agent optimization — prompt scaffolds, subagent orchestration, memory and planning patterns that hold up across long sessions.
Eval harnesses for my own workflows — measuring agent reliability on real tasks instead of trusting vibes.
Full-stack scratch builds (React, Node, Flask) and a few distributed-systems holdovers from my indie-dev stretch — smaller scope, useful for thinking.

ask_logan · mcp

ask_logan — the chat in messages — also runs as a local Model Context Protocol server. Attach it to Claude Desktop or any MCP host to call the same three tools from your own agent.

send_contact_email — drafts an email to me
lookup_patent — semantic search across 90 filings
link_to_page · recommend_next_page · get_career_timeline · compare_engagements

~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "ask_logan": {
      "command": "node",
      "args": ["/absolute/path/to/personal_website/mcp-server/server.js"]
    }
  }
}

see how this is built →

open to

Solutions-architect, enterprise-architect, and AI-platform roles where the work is real — production systems, governed AI, agentic depth, integration depth. Open to conversations on workshops, architecture reviews, invention disclosures, and senior IC paths.

building toward the next layer of enterprise AIby shipping, learning, and showing up.

change log