← all challenges

// challenge brief · hack a ton 2026

Five Star

A HoReCa owner onboards their business and gets personalised, prioritised tips to raise the quality of their service.

// Proposed by
Ambasada
// Industry
HoReCa / hospitality
// Difficulty
🟢 Beginner-friendly

// the problem

A seasonal coast lives and dies on guest experience. A beach hotel, a seafront restaurant, a city café, a terrace bar: each one has the same truth sitting in plain sight inside hundreds of reviews, scattered, multilingual, and never read systematically. Owners skim the latest few and miss the pattern. None of them has a tool for it, so service quality across the whole coast stays stuck wherever the loudest few reviews left it.

// your mission

Build a platform a HoReCa owner self-onboards onto, brings in their reviews, and gets personalised tips on how to improve the quality of their service. The same engine works for a hotel, a restaurant, a café, or a bar. Enrichment is encouraged: the more relevant context you bring in about the business, the more specific and actionable the tips.

// how it works

  1. 1Onboard: the owner registers their business (type: hotel / restaurant / café / bar) and brings in their reviews (bring your own, or use a public HoReCa review dataset; for the live demo, a partner business's own reviews are the strongest signal).
  2. 2Extract: per review, capture sentiment, the specific aspect, severity, and language. The relevant aspects depend on the business type (a café has no pool; a hotel has no menu):
    • Food & drink: food quality, taste, freshness, portion size, menu range, coffee/bar.
    • Service: speed, staff friendliness, order accuracy, waiting time, check-in/reservation.
    • Place: cleanliness, ambiance/atmosphere, noise, AC, Wi-Fi, pool, location.
    • Value: price / value for money.
  3. 3Enrich: go beyond the raw review text. Pull in whatever extra context sharpens the advice for this business: its type, menu, photos, location, the season. This is where good solutions pull ahead.
  4. 4Rank: cluster issues and order them by frequency × severity × recent trend.
  5. 5Advise: output a personalised "fix this first" list, tone-matched multilingual draft replies, and a weekly digest. Optionally push tickets via webhook (human approves first).

// suggested approach

  • Inference: for model access, we suggest [LLMok](https://llmok.app). Use code AMBASADA26 for 50% off.
  • An LLM for aspect/sentiment extraction and clustering; a simple dashboard for output.
  • Multi-tenant note: the business type should steer which aspects matter and the reply tone. The same pipeline, parameterized per vertical, beats one hotel-only model.
  • Enrichment is encouraged: the goal isn't summarising reviews, it's turning them into advice specific to one business. Bringing in extra context (the menu, the photos, the location and season, the owner's previous fixes) makes a tip concrete rather than generic.
  • Floor version: one business, its reviews, extract aspects + sentiment, output the ranked tip list. Add enrichment, a second business type, replies, and trends from there.

// how you'll be judged

  • Accuracy of aspect/sentiment extraction.
  • Usefulness and personalisation of the tips. Would an owner actually act on them?
  • Enrichment: how much relevant context the system brings in to make the advice specific.
  • Works across HoReCa types, not just hotels.
  • Onboarding: how easily could a real owner sign up and get value with their own reviews?
  • Reply quality and multilingual correctness.
  • Speed / scalability over many reviews and many businesses.

// stretch goals

  • Time-trend detection ("AC complaints spiked this month").
  • Fake/incentivized-review flagging.
  • Competitor comparison.
  • Cross-business benchmarking on the platform ("your service speed vs other coast restaurants").

// deliverables

  • Git repo: code + README + run instructions.
  • Demo: a real HoReCa business's weak spots, ranked, on screen (show the owner onboarding flow if built).
  • 5-minute pitch: architecture, decisions, trade-offs.