EXCAR is pre-seed and building in the open. This page separates four things most decks blur together: what runs today, what we're verifying right now, the next gate we have to clear, and the long-term product vision. If you'd like the deck or a conversation, details are at the bottom.
EXCAR is an AI voice companion that gives the car you already own a personality, memory, and an understanding of the road.
Automakers are shipping capable assistants, but only in new vehicles, tied to one brand's dashboard. Meanwhile there are 289 million vehicles on U.S. roads and the average one is 12.8 years old. Those drivers get a phone mount and a charging cable.
A phone on a mount isn't a solution either. It can't see the car, it wasn't designed for a moving cabin, and it asks the driver to look down at exactly the wrong moment.
No dashboard surgery, no brand lock-in, no new car.
The fastest way to lose an investor's trust is to present a plan as a result. So we don't.
We publish no completion percentages, no unit sales, no partnerships and no certifications, because we have none of those yet. What we have is a device that works and a verification record that says how well.
Seven products are on our map. One is in development. That gap is a decision, not a gap in the plan.
At any moment: one consumer hero product, one low-cost experiment, and at most one paid B2B pilot. Never more.
| Horizon | Consumer | Professional and B2B | What has to be true to move on |
|---|---|---|---|
| Now | EXCAR One in development | Fleet and dealer discovery conversations | Voice quality in a noisy cabin, secure vehicle link, someone paying for it |
| Next | EXCAR One controlled launch | One paid rideshare pilot, three fleet design partners | Gross margin, returns, retention, measurable pilot ROI |
| Then | EXCAR Air cost-down decision | Fleet v1, certified dealer installs | Real One usage data showing what can be removed safely |
| Later | EXCAR Buddy as a limited experiment | EXCAR Mirror validation, OEM conversations | A platform partner, and a compliance path we can afford |
Hardware companies rarely die of a small vision. They die of two product programmes at once: two bills of materials, two certifications, two supply chains, half the attention on each.
The expensive parts of EXCAR, the voice engine, the memory, the vehicle data layer and the app, are written once and reused. So the second product is a housing and a channel decision, not a second company.
Everything beyond EXCAR One is a direction, not a commitment. Nothing else is scheduled, priced or staffed.
The existing fleet, not the new-car lot, is where the next few years of in-car AI actually get decided.
Source: S&P Global Mobility, 2025. U.S. light-vehicle average age: 12.8 years; vehicles in operation: 289M.
Source: Meticulous Research, Automotive Voice AI Assistants Market, 2026–2036. $2.87B to $18.92B; 21.3% CAGR.
For context, broader industry estimates point the same direction, though scope and methodology vary by firm: Grand View Research expects the global connected car market to grow from about $12.8B (2024) to $26.5B by 2030 (12.8% CAGR), and 2025 industry reporting on aftermarket vehicle telematics hardware puts global shipments at roughly 51 million devices in 2024, expected to reach about 77.5 million by 2029, a sign the retrofit hardware category itself is already large and growing, independent of voice AI specifically. These are general market signals, not EXCAR-specific forecasts.
Automakers are shipping AI assistants, but only in new vehicles. The overwhelming majority of cars on the road, and virtually the entire used-car market, will never get a factory upgrade. That installed base is our market: one device, one plug, and any car gains a modern AI voice companion.
The product is aftermarket-simple and assistant-complete: music, navigation handoff, a briefing on the way to work, and live vehicle health, reading real fault codes and explaining them in plain language. It carries context from one drive to the next, in one of the most consistent, captive, single-user environments a person spends time in.
Our first target is model year 2008 and newer. From 2008, every new vehicle sold in the U.S. was required to expose OBD-II over a standardised CAN interface. That single decision is why one device can speak to cars from dozens of brands without custom wiring, and it is where we put the first EXCARs on the road.
Three categories already talk to drivers. None of them is built for the car sitting in the driveway right now.
There is also a fourth group worth naming: aftermarket OBD dongles and screen-mirroring adapters. They add a readout or a display. They don't add a companion, and they have no reason to know who you are.
EXCAR grows outward from a single, tightly-supported market before scaling nationally.
Real driving usage and structured feedback in EXCAR's home market.
Controlled testing with people close to the team.
Community and campus partnerships across the region.
Technology and automotive early adopters.
Bringing EXCAR One to drivers across the United States, backed by pilot data and a proven install-and-support model. Only then does a second product become a question worth asking.
A one-time device purchase gets EXCAR into the car. A monthly subscription covers the intelligence, memory and updates that make people keep it there. Pricing is not yet finalised, and we won't publish a number we can't stand behind.
EXCAR One for the cabin, with EXCAR Link for the port. Priced once the custom board tells us what a unit truly costs to build.
AI access, personalisation, long-term memory and updates. The recurring line, and the one that compounds as the relationship deepens.
Pre-seed capital at this stage buys exactly three things: closing engineering unknowns, building measurable pilot units, and the compute to train and evaluate our own model. Not marketing, not headcount, not an office.
Our own board is also what makes the unit economics real. Until that exists, any margin figure would be a guess, and we'd rather tell you that than dress one up.
Full voice pipeline running in a real vehicle: wake word, streaming conversation, music control, navigation handoff, and a companion iOS app on TestFlight.
Moving to more capable on-board hardware for radar and faster response times, while we train EXCAR LLM, our own model, from strong open weights for conversation in the car, building directly on Dr. Atilla Kaan Alkan's doctoral research in NLP and deep learning at Université Paris-Saclay, so the heart of the product stays in our hands.
Hands-on installations with real drivers to validate retention, latency targets, and willingness to pay for the hardware-plus-subscription model.
Raising with pilot data, retention numbers, and a defined hardware supply chain in place.
Pre-orders to the broader market, backed by real in-vehicle data and a working manufacturing path. We're not dating this one, because it depends on how the pilots behave.
Four inputs. One verifiable outcome. If a proposal doesn't move the centre of this diagram, we're not spending on it.
Ten drivers, ten cars, installed by us, with latency, retention and failure modes measured rather than described.
Train and evaluate EXCAR LLM, and build the in-car evaluation set the model has to pass.
Build EXCAR units that are measured and repeatable, so ten cars produce comparable data instead of ten anecdotes.
Close the electrical, mechanical, thermal and acoustic unknowns with people who have closed them before.
Connect AI development to automotive manufacturing, integration and certification. The part money alone doesn't buy.
We'd rather have an investor who opens one automotive door than one who wires twice the cheque and disappears.
Capital gets us to the pilot. Introductions to automotive engineering, manufacturing, integration and certification are what turn a validated pilot into a product that ships. If that's the kind of investor you are, this is the stage where you matter most.
Business, engineering and AI research, with no outsourced core.
We didn't start as a team looking for an idea. We assembled ourselves piece by piece around one: a product person who wouldn't let the idea go, an engineer who could actually put it in a car, and a language-model researcher who could give it something worth saying. Each of us was the missing piece the other two couldn't work around, and none of us joined until the shape of the thing was clear. That's why the core is built in-house, and why nothing important is outsourced.
Product strategy, fundraising, partnerships and go-to-market. Owns the question of who EXCAR is for and what they'll pay for it.
The end-to-end technical stack: vehicle hardware integration, the real-time voice pipeline, mobile and connectivity. If it listens, responds or connects, he built it.
Doctoral research in NLP and deep learning at Université Paris-Saclay, now a postdoctoral researcher at Harvard. Leads the conversational intelligence and the EXCAR LLM effort.
The relevant experience is narrow and we'd rather say so: deep in language modelling and real-time systems, deliberately buying in automotive hardware expertise where we don't have it. That gap is the first item on the ask above.
We're a pre-seed team building in public. Drop your email and we'll send the full deck and financial model, or reach out directly and a founder will respond.
