← Back/Projects / Software and AI

Applied AI and Automation

Builder, Workflows, agents, and tooling, 2025 to present

ProblemSmall teams drown in repeatable work: call prep, memos, follow-ups, CRM hygiene.
My roleDesigned and built the workflows and the frameworks behind them, end to end.
ResultTCP’s first AI operations stack, reusable skills for call prep, IC packages, and outreach, and a personal second-brain pipeline that is growing into a closed loop.
LinksNone public yet
RoleBuilder
ContextWorkflows, agents, and tooling
Dates2025 to present
FieldApplied AI, Workflow Automation
SkillsClaude Code, Agentic Workflows, Multi-Agent Systems, Retrieval and Second-Brain Systems, Integration (Notion, HubSpot, Drive, Calendar), Framework Design, Python, Full-Stack Development, Scrapers and APIs, Prompt Engineering

What I build

  • An AI operations stack for Tulypp Capital Partners across Claude, Notion, HubSpot, and Google Drive, with reusable skills for call prep documents, investment committee packages, and outreach
  • Multi-agent workflows that research a company, draft a brief, and hand off a formatted document, with a human review step before anything leaves the building
  • Personal tooling: a morning brief wired to my calendar, task automations, and the full-stack habits that produced ToleranceForge, where a multi-model pipeline reads drawings and prices tolerances
  • A second brain. A Telegram-to-Obsidian pipeline turns video links into transcribed notes and then into concrete proposals for my own workflow. Capture is mechanical and spends no model tokens on judgment. Ideation runs only when I ask. It is private and still a work in progress, but it is the system I actually use

Where it is going

The pieces are being wired into one loop. The calendar is already connected. Next are calorie tracking, WHOOP, and gym logs, so the same system holds a full health and wellness profile and can suggest changes from it. The second brain will take articles, news, documentaries, and videos, store what matters, and learn with me. From there it should extract what is useful and turn it into updates to the tools themselves, a closed loop where the system helps build its own next version. The end state ties into Onshape as well, so design work gets the same treatment. Call it the full JARVIS setup. I am building it piece by piece.

What kind of AI work this is

Agentic workflows with tool use and a human review gate. Retrieval over a personal knowledge base. Multi-model pipelines where each model does the step it is best at. Skill and prompt frameworks that other people on the team can run without me. Most of it is framework design more than model building: deciding what gets automated, where the human sits, and what the system is never allowed to do on its own.

Specifics about internal firm workflows stay off this page. This card exists to show the pattern: find the repeatable work, build the tool, keep a human in the loop.