Hi,
I'm Paul Sin.
Creating AI-Driven Systems and Full-Stack Experiences that think, adapt, and scale.

The agent generates multiple thumbnail options, scores them with a small eval loop, and then returns the top pick with title, layout notes, and prompt references so designers or other tools can ship assets consistently.

Pipeline parses the transcript, ranks highlight moments, drafts titles and descriptions, and suggests music or sound effects so creators can push Shorts faster with less manual editing.

The graph classifies each message, retrieves the right policy or FAQ content, drafts a response, and escalates tricky cases with a clean summary so human agents can jump in without re-reading the whole thread.

The flow reads incoming tickets, detects the type of request, fills a brief with key details, and posts updates to Slack and Notion so PMs and stakeholders always see the latest status without asking.

Webhook events drop into a LangGraph workflow that cleans data, calls external APIs, requests approvals, and ships notifications. Logs and simple metrics help debug runs and keep an eye on failures.

Agents work together to read the user’s profile, search job boards, suggest matches, and draft customized resumes and emails. A lightweight memory layer remembers preferences like tech stack, location, and salary range.