Our client is a German automotive technology company selling products and services for dealerships across Europe. Our challenge: turn five years of untouched pricing data into real-time recommendations that help dealers stop guessing.
Automotive dealerships set vehicle prices based on gut feel. Our client had five years of pricing history sitting across their dealer network.
The result was predictable: vehicles overpriced and stalled on the lot, or sold too quickly, leaving margin behind. Our client wanted to fix that by offering AI-powered pricing recommendations to their clients.
The PoC had to prove that Amazon SageMaker could train on the client's real vehicle data, expose the model through a production-ready API, and return confidence-scored recommendations.
We built an AWS-native ML pipeline that ingests raw vehicle data and returns actionable pricing recommendations. We focused on:
A data pipeline that cleans the data first: AWS Glue extracts pricing history and vehicle specifications from client's systems, engineers the relevant features, and loads a clean training dataset into S3 — with 100% data integrity on critical fields.
A model trained on real dealership behaviour: Amazon SageMaker trains on five years of client's vehicle data. It learns which characteristics actually drive price differences across vehicle types, and generates feature importance analysis to make recommendations explainable.
Real-time inference: A Lambda function processes each request, calls the SageMaker endpoint, and returns a structured pricing recommendation with a confidence score — in under 500ms for 95% of requests.
A feedback loop for continuous improvement: A second Lambda captures whether each recommendation was accepted or overridden, storing dealer responses in RDS PostgreSQL to feed future retraining cycles.
All infrastructure is provisioned via Terraform for repeatable deployment and clean handover.
The PoC validated the full pipeline in three weeks. Our client now has a working ML system that trains on five years of vehicle data, differentiates pricing patterns across vehicle types, and serves low-latency recommendations through a production-ready API.
Our client plans to integrate the pricing engine into their platform as a core product feature, and offer it as a B2B API to dealer management system providers, extending their reach beyond display hardware.
Tell us more about yourself and what you're got in mind.
Prefer email? hello@apexlab.io
Or even a quick call?
We don’t care about fancy CVs or long interview processes. Provide some basic information, let’s have a quick chat, and then show us what you love doing the most: write some code and create a product.
