Hackathon Winner

Autonomous Marketplace Flipper Agent

First-Place Winner – Autonomous Marketplace Flipper— AI Tinkerers SF “Agents That Buy & Sell” (LiquidMetal & Vercel) AI Tinkerers

Event & Result: Won 1st place at AI Tinkerers San Francisco’s mini-hack Agents That Buy & Sell on Oct 1, 2025 (hosted with LiquidMetal AI & Vercel). Teams built marketplace agents—buyers, sellers, or brokers—using LiquidMetal’s Raindrop components (SmartBuckets, event triggers, agent memory).

What We Built (at a glance):

  • Broker agent that scans new listings, estimates fair value, negotiates price, and flips inventory for profit.
  • SmartBuckets to store/search listings, offers, and counter-offers; event triggers to act instantly when new items hit the market; agent memory to adapt strategy over time.
  • Buyer/Seller micro-agents with roles:
    • Buyer → valuation + budget guardrails + auto-negotiation
    • Seller → dynamic pricing + relist logic + reserve rules
    • Broker → matches buyers/sellers, arbitrages spread, settles deals

Demo Highlights:

  • End-to-end loop: discover → value → negotiate → settle → relist (profit tracking).
  • Negotiation policy that adjusts concession rates based on time-since-listing and observed counter-party behavior.
  • Safety rails (max loss per trade, cooldown after failed negotiations).

Why it won (judges’ angle):

  • Clear market behavior (buy/sell/relists) aligned tightly with hack theme.
  • Practical use of LiquidMetal Raindrop primitives (SmartBuckets, triggers, memory) instead of “toy” agents.
  • Fast, legible UX for live demo (deal log + P&L over session).

Tech Stack:

  • Agents/Infra: LiquidMetal Raindrop (SmartBuckets, event triggers, agent memory).
  • Runtime: Python (valuation models, negotiation policy), small Node/Express (webhook/event relay).
  • UI: Lightweight Vercel app for live deals feed and P&L.
  • Data: JSON listings; cached features (historical price, time on market, demand score).

My Contributions:

  • Implemented valuation model + concession policy (time/price-elastic negotiation).
  • Designed deal state machine and fail-safe rules (budget caps, cooldowns).
  • Built the demo dashboard (offer stream, status, profit).

Links:

Other Projects