Work
Not a portfolio wall. A set of working AI-native organization samples.
Five products. One operator. Each product proves a direction: turning fuzzy human intent into systems that agents can execute, inspect, and reuse.
One Page
Focus in one gesture.
- Problem: Meetings, ideas, and tasks scatter across tools. The user needs one low-friction place to recover focus.
- Agent role: Turn raw input into actions, context, and next steps instead of only producing polished text.
- Proof signal: The product shifts a note from “recorded” to “ready to act on.”
MARGIN
Natural language → Action.
- Problem: People express intent in language, while traditional software asks them to adapt to fields and flows.
- Agent role: Parse the action, object, and constraints from a sentence, then route them into an executable workflow.
- Proof signal: The user does not learn a complex interface; the system turns one sentence into a next move.
FileFlow
Safe automation.
- Problem: File automation is dangerous when it moves fast and nobody can see what changed.
- Agent role: Split bulk work into auditable steps while keeping human approval at the important boundaries.
- Proof signal: Automation becomes inspectable, bounded, and reversible instead of a black box.
Fairmate
Your negotiation team.
- Problem: One person negotiating alone has to hold strategy, materials, objections, and follow-up at the same time.
- Agent role: Simulate a small team that prepares arguments, anticipates objections, and organizes next-round actions.
- Proof signal: It is not a chat assistant. It is a set of roles working around a negotiation goal.
SFA
AI sales coach.
- Problem: Sales training does not need more generic scripts. It needs repeated practice close to real business context.
- Agent role: Distill meeting recordings, business knowledge, and sales methodology into specs, then let an agent team build in parallel.
- Proof signal: 52 development specs, 2,716 passing tests, and 105,000 lines of code delivered in 4 days.
Building in this direction? I'm reachable →
