virtual showroom

Inspiration: We've all been there — you walk into a bike showroom, a salesperson throws 40 models at you, you have no idea what half the specs mean, and you leave more confused than when you came in. The problem isn't the bikes. It's that nobody asks you the right questions first. What's your height? Your budget? Are you commuting through city traffic or hitting mountain trails on weekends? We wanted to build the showroom experience that should have existed all along — one that listens before it sells.

What it does: virtual showroom is a smart bike recommendation engine that covers both motorcycles and bicycles. You answer 5 simple questions — your budget, your height, your riding purpose (commute, sport, off-road, or leisure), your experience level, and your preferred bike type — and virtual showroom scores and ranks the best matches from a dataset of 500+ bikes. It doesn't just show you a filtered list. It explains why each bike fits you — "this one matches your height range, sits within budget, and is rated beginner-friendly for city commuting." The top 3 recommendations come with full specs, a match score, and similar alternatives in case you want to explore. It's deployed as both an interactive web app anyone can use and a live API that developers can integrate into their own platforms.

How we built it: We started by identifying the core user inputs that actually determine bike fit — height, budget, purpose, and experience. From there we sourced and merged multiple public datasets covering motorcycle specs (engine cc, seat height, weight, price, category) and bicycle specs (type, frame size, weight, terrain, price range). The raw data was messy — inconsistent column names, missing values, mixed units — so we used Zerve's AI agent to write and iterate on the cleaning pipeline, which saved hours of manual work. Once the dataset was clean, we designed a weighted scoring model that ranks bikes based on how well they match each user input. Purpose and experience level carry the most weight; budget and height act as hard filters. We then built the recommendation logic, tested it across edge cases, and deployed the whole thing as an app and API directly from Zerve without ever leaving the platform.

Challenges we ran into: The biggest challenge was data. There's no single clean dataset that covers both motorcycles and bicycles with consistent, comparable fields. We had to source from multiple places, standardize units (cm vs inches for seat height, USD vs local currencies for price), and create a unified schema that worked for both bike types despite them being very different products. Another challenge was the scoring model itself — making it feel smart rather than mechanical. A pure filter just returns a list. We wanted the engine to make a genuine recommendation with reasoning, which meant designing the weighting logic carefully and testing it with real-world scenarios. We also had to think hard about edge cases: what if someone is very tall with a very low budget? What if a beginner wants a sport bike? The model needed to handle those gracefully.

Accomplishments that we're proud of: A lot of recommendation engines are black boxes — you get a result but no reason. virtual showroom tells you exactly why it picked each bike, which builds trust and actually helps people learn what specs matter for their riding style. We're also proud of the dual deployment — shipping both an interactive app and a live API from the same project means virtual showroom isn't just a hackathon demo. It's something developers can build on and users can actually use today. Finally, pulling off a clean cross-category recommendation engine (motorcycles AND bicycles in one model) with a unified scoring system was harder than it sounds, and we're happy with how it turned out.

What we learned: The most important thing we learned is that the quality of your question determines the quality of your output. Before we wrote a single line of code, we spent time just thinking about what inputs actually predict a good bike match. That upfront thinking made everything downstream — the data model, the scoring logic, the UI — much cleaner. We also learned how powerful it is to have an AI agent handle execution while you focus on direction. Zerve let us stay in strategy mode: deciding what the model should prioritize, what edge cases to handle, what the output should feel like. The AI handled the code, the debugging, and the iteration. That division of labor is genuinely the future of building.

What's next for Virtual Showroom: The recommendation engine is the foundation, but the vision is much bigger. Next we want to add 3D bike visualization so users can actually see what their matched bike looks like from every angle — the full virtual showroom experience. After that, live dealer inventory integration so virtual showroom can tell you not just what bike fits you but where you can buy or test ride it near you. Further down the road: test ride booking directly through the app, user reviews and ownership insights layered into the recommendation, and expanding the platform beyond bikes to scooters, cars, and other vehicles. The long-term goal is to make virtual showroom the starting point for every vehicle purchase decision.