← all challenges

// challenge brief · hack a ton 2026

Freight Broker

No single carrier ships your container from China to Constanța. This agent finds the cheapest way, leg by leg.

// Proposed by
Ambasada
// Industry
Maritime logistics
// Difficulty
🟢 Beginner-friendly

// the problem

A port operator's real complaint: when you buy goods from China, no single transporter can ship the whole journey. The vision is a freight exchange where operators bid on portions of the trip. The question that matters: what's the cheapest end-to-end route once you combine those bids?

// your mission

Build a freight-broker agent. A shipper describes a need in plain language; the agent breaks the journey into legs, assembles the cheapest valid end-to-end route from available carrier bids, and drafts the bookings.

Example: "2 containers of furniture, Shenzhen → Constanța → Cluj, within 6 weeks."

// how it works

  1. 1Parse: turn the free-text request into structured legs (cargo, origin, destination, constraints).
  2. 2Look up: pull carrier bids per leg (sea, port handling, rail, road) from a bid dataset you assemble (a small sample or synthetic set is fine).
  3. 3Optimize: find the cheapest end-to-end route honoring capacity and time.
  4. 4Act: present the route with a per-leg cost/time breakdown and draft a booking message to each winning carrier.

// suggested approach

  • Inference: for model access, we suggest [LLMok](https://llmok.app). Use code AMBASADA26 for 50% off.
  • LLM to parse the request into structured fields.
  • Cheapest-path over a small graph of legs, using Dijkstra or even brute force. No optimization theory required.
  • LLM to draft the booking messages.
  • Floor version: parse a fixed-format request, find the cheapest path ignoring constraints, show the breakdown. Add capacity/time constraints and booking drafts next.

// how you'll be judged

  • Correctness of the cheapest route.
  • Constraint handling (capacity, transit time).
  • Quality of the free-text parsing and the booking drafts.
  • Robustness on tricky or under-specified requests.
  • Creativity, such as cost-vs-time trade-offs and simple bid negotiation.

// stretch goals

  • Multi-objective routing (let the user weight cost vs. speed).
  • A negotiation loop with carrier agents to drive prices down.
  • Live re-routing when a leg becomes unavailable.

// deliverables

  • Git repo: code + README + run instructions.
  • Demo: a request resolved into a priced, booked route on your own sample data.
  • 5-minute pitch: architecture, decisions, trade-offs.