Overview
Carvexa is a SaaS ERP and market-intelligence platform for used-car dealerships. It covers the operational flow of a dealership end to end: vehicle intake, stock, valuation, margin analysis, documentation, AI-generated vehicle reports and automated listing publication. More than 100 sources are collected, normalized and analyzed to surface market signals and proprietary KPIs such as expected margin, estimated days on market and independent vehicle valuation. The company is currently being formed and the product is being prepared for its first customers.
An ERP and market intelligence layer for used-car dealers
Carvexa is designed around the complete operating cycle of a used-car dealership: receiving a vehicle, enriching its data, valuing it, deciding whether to buy it, managing it in stock, preparing its documentation and publishing it for sale. The product combines those workflows with a live market view, so purchasing, pricing, margin and vehicle-rotation decisions can be made from the same application instead of through disconnected tools.
A near-complete catalogue of the Spanish used-car market
The platform aggregates more than 100 sources, including listings, DGT data, reports and vehicle catalogues. These sources are normalized into a broad historical catalogue covering more than 95% of vehicles, with detailed information about versions, prices, optional equipment and colours. This foundation makes it possible to compare a vehicle against a much wider market instead of evaluating it in isolation.
A distributed system built to process the market continuously
The collection pipeline is coordinated by an orchestrator that sends work to four slave servers, with a fifth main server responsible for coordination and core services. Proxies, queues, Redis and containerized workers allow the system to distribute requests, control source-specific limits and recover from partial failures. Built with Python and TypeScript, the pipeline is optimized for bursts of more than 50,000 listings per hour from over ten active sources. At that throughput, the complete Spanish second-hand market can be reviewed in under 24 hours, enabling metrics such as average time on market, price evolution and expected margin to be recalculated frequently.
From raw listings to useful signals for a dealership
Carvexa turns collected data into operational intelligence: proprietary vehicle appraisal, expected margin, comparable prices, market position and estimated days on sale. It also analyzes listing images with OCR and an in-house model to identify interior wear, paint damage and other visible defects. Combining structured catalogue data with visual evidence supports more consistent valuations and better purchasing decisions. The result is a vehicle report generated automatically with AI, combining the catalogue, market comparables, image findings and the relevant context for the dealership.
Reports, listings and lead qualification in the same workflow
The application uses RAG, multiple specialized agents and Typst to turn structured vehicle data and unstructured source material into consistent, technical reports. Once a vehicle is ready for sale, the system can generate a complete listing — title, description, equipment, highlights and pricing context — and publish it automatically to the configured portals. It also responds to incoming buyer messages and applies an AI lead funnel: it interprets the conversation, identifies buying intent and filters out low-quality or purely exploratory contacts so the dealership team receives leads with a minimum level of genuine interest. The dealership can review the result, keep control of the final action and avoid rewriting the same information for every channel.
A data platform designed for throughput and observability
The platform runs as a Dockerized system across five servers: one main node and four slave nodes. PostgreSQL stores the operational and normalized catalogue data, MongoDB supports document-oriented workloads, Redis handles fast state and coordination, and Grafana provides visibility into processing, errors, throughput and system health. This separation makes it possible to scale the expensive collection and analysis work independently from the core dealership application.
Cofounder leading development and data
I am cofounder and lead the development and data work across the product. My responsibility covers the architecture of the collection system, data pipelines, catalogue construction, processing optimization and the software layer that turns the resulting intelligence into a usable dealership product. Carvexa is currently in company-formation phase and looking for its first customers.