The trips we carry are what will teach the model that plans the next one.
Tour operators sell complex multi-stop journeys with tooling built for hotel rooms. We build the layer underneath: a model that plans the trip, the surfaces that carry it, and an arena that scores what comes back. The score trains the model. Each turn of that loop makes the next trip better and the layer harder to replace.
The loop is the product.
The stack is the company.
Everything in the loop below reads one engine: one itinerary document, one map, one brand record. Once that exists, two things become possible: a model that plans on it, and an arena that scores it.
An operator does not buy a website from us. They keep their own brand, their own traffic, their own customer.
One document, six live surfaces.
Pick a turn. The real prototype loads on the right, not a screenshot of it.Its vocabulary is places, not words.
A language model learns which words follow which. This one learns which places follow which, from 56,731 real venues sitting at their real coordinates.
In the catalog, every venue is a bookable record, and its own embeddings fell into eight semantic clusters, not categories anyone typed in.
{{ ltm.briefLabel }}
LLM’s write a piece of text, and hallucinate
A travel writer who’s never been on a trip
{{ ltm.emptyText }}
Every distinct piece is a node, placed on the sphere and given one of the eight hues by a hash of itself. Neither position nor colour means anything. A wire joins each piece to the next; a node fires when its piece arrives.
{{ ltm.frontierNote }}
Marco Polo makes a decision, deterministically
Connect with a specialist for a trip you created
{{ ltm.emptyEvents }}
Every place is a node on the earth at its real coordinates, on its real coastline. The route is the wire. A place fires when its event arrives. After TRIP END a light travels the route, hub by hub; that clock is presentation, the coordinates are the record.
{{ ltm.mapNote }} {{ ltm.engineNote }}
Left, text checked against the world afterwards. Right, entries of a catalog, real by construction. {{ ltm.closing }}
Wayfare measures. MarcoPolo competes.
Every planner claims it plans real trips. Wayfare is the definition of what a plan must survive to count, made runnable, and it gates every release of our own engine.
{{ arena.zoomDesc }}
The score is the reward. The arena gates each release of the engine, and its reward signal is the verifiers plus one outcome label, no_edit@24h: a traveller who left the itinerary alone for a day. That is the outcome the surfaces above are built to produce, which is where the loop closes. Back to the engine ›
Custom planning across the travel industry is still archaic.We built the conversion layer it should have had over a decade ago.
Tech companies know how to build product, but most underestimate how messy travel really is: fragmented suppliers, custom pricing, human sales processes, offline inventory, destination nuance, and the gap between what a traveler wants and what a business can actually deliver.
tripcreator.ai exists because Marc has lived both sides of that gap for nearly two decades, as a DMC partner, B2B travel SaaS operator, marketplace executive, and global partnership leader. With Edoardo as CTO, the vision finally has the technical architecture to match the industry problem: AI that is not just impressive in a demo, but grounded in real travel inventory, real business workflows, and real conversion.
18 years as a partner at a Latin American DMC, scaling it from $3M to $11M across eight regional offices. Partner in Itereo, a B2B custom trip planning SaaS, since 2013. Most recently, VP of Partnerships and Sales at one of the largest custom-trip marketplaces in the industry, leading global partnerships with enterprise tour operator brands and tourism boards across multiple continents. The problems tripcreator.ai solves aren’t theoretical. He’s lived the gap between traveler intent and business delivery on every side of the industry.
Leads AI architecture and engineering. Deep expertise in neural network systems, high-performance backends, and enterprise cloud infrastructure (GCP, Azure). Previously a Solution Architect Lead in AI at OverIT. Edoardo’s focus is on where AI in travel is going: not the current generation of LLM wrappers, but the deterministic, domain-specific intelligence layer that the industry surfaces demand for making it easy. tripcreator.ai is his bet on what the team’s vision looks like, built right, from the ground up.
Travel needsan intelligence layer.Not another chatbot.
Someone arrives with an idea. A tour operator, a destination, a host agency, an event organizer: each one faces the same challenge, to translate vague human desire into something structured, feasible, and deliverable. The gap between what a traveler imagines and what a business can actually offer costs everyone time, money, and opportunity. It’s not a search problem. It’s an intent-to-action problem.
AI tools have flooded the space promising to fix this, but most of them work against real businesses instead of with them. They hallucinate inventory that doesn’t exist, generate plans no one can deliver, and hand the traveler somewhere else to book. tripcreator.ai is built the other way around: it starts with what the business actually offers, learns from what travelers genuinely respond to, and closes the loop on the business side, not the consumer side.
“Bridging that gap between what’s in someone’s head and turning that into reality for us in the background.”
Travel is the wedge.The workflow is the company.
We’re building on travel because custom travel is the clearest, highest-friction expression of a workflow problem that shows up everywhere experiences are sold: a person arrives with vague intent, a business has to translate that intent into a structured, feasible, bookable plan, and the gap between those two states burns weeks of margin on every transaction.
That same workflow lives inside event management, conference programming, multi-day festivals, destination weddings, and curated corporate retreats: anywhere a buyer is shaping a personalized experience against real-world inventory, time constraints, and operator rules.
For the next eighteen months, we’re focused on travel, proving the engine in the hardest, most fragmented expression of the workflow. Adjacent verticals follow when the engine is mature, the unit economics are proven, and the right industry partner walks through the door.
“Teaching machines how humans discover meaning through exploration.”
Let’s build somethingworth traveling for.
We work with operators, investors, and partners who believe travel deserves better infrastructure. If that’s you, we’d like to talk.
Private demonstration · Sample partner brands shown with permission · Not for distribution
marc@tripcreator.ai ›