Executive Summary:
As the connected-vehicle fleet grew, the e-mobility startup needed faster telemetry, better battery insight and a practical way to predict maintenance issues. Rapyder built the AWS AI and machine learning foundation behind it: secure IoT ingestion, Kinesis streaming, an S3 data lake, SageMaker models and an MLOps pipeline with staged release and human approval. The platform improved telemetry visibility by 60–70%, accelerated processing by 50–60% and gives analytics room to scale with the fleet.
About Customer:
The customer is an e-mobility startup building modular, customisable electric two-wheelers. Its connected vehicles generate data on battery performance, diagnostics, energy consumption and operational health to support fleet monitoring and improve the ownership experience.
Problem Statement:
The e-mobility startup lacked timely visibility into vehicle performance and struggled to process high-frequency telemetry as its fleet expanded. Gaps in predictive maintenance increased service costs and unplanned downtime, while uncertainty around battery degradation affected customer confidence. Fragmented analytics workflows made it harder to turn vehicle data into timely operational decisions.
Industry:
Industries, Others
Offering:
Offerings, Data Analytics
Business Challenges:
- Visibility gap: Limited visibility into vehicle performance reduced the operational insights available to the team as the connected vehicle fleet grew.
- Telemetry scale: The platform was unable to process high-frequency telemetry at scale, delaying decision-making across operations.
- Predictive maintenance gap: A lack of predictive maintenance capabilities increased service costs and unplanned vehicle downtime.
- Battery uncertainty: Uncertainty around battery degradation reduced customer confidence in vehicle performance and lifecycle expectations.
- Fragmented workflows: Fragmented analytics workflows slowed innovation cycles as the company scaled its connected vehicle initiatives.
Solution Implemented:
- Telemetry ingestion: Built a telemetry ingestion layer where AWS IoT Core securely receives vehicle telemetry through MQTT-based communication channels, with each connected vehicle provisioned using X.509 certificates and governed by fine-grained IoT security policies.
- Streaming pipeline: Routed telemetry streams through Amazon Kinesis Data Streams to enable high-throughput processing and real-time analytics workflows.
- Event processing: Used AWS Lambda to process incoming events, perform validation and anomaly detection, and orchestrate downstream workflows.
- Data lake: Established Amazon S3 as the centralized data lake for raw and processed telemetry data, with Amazon DynamoDB providing low-latency access for operational applications.
- SageMaker models: Deployed Amazon SageMaker to power model development, training, deployment, and continuous retraining for battery state-of-health prediction, vehicle range estimation, energy consumption optimization, predictive maintenance, and performance analytics.
- MLOps pipeline: Built an automated MLOps pipeline covering model build (data pre-processing, training, and evaluation via SageMaker Pipelines) and model deployment (staging, manual approval, and production release via CodePipeline and CloudFormation).
- Security controls: Implemented device, data, and operational security controls, including X.509 certificate-based device authentication, encryption in transit and at rest, IAM least-privilege access, and continuous monitoring via CloudWatch.
Technology Stack:
- AWS IoT Core
- Amazon Kinesis
- AWS Lambda
- Amazon DynamoDB
- Amazon S3
- Amazon SageMaker
- AWS CodeBuild
- AWS CodePipeline
- Amazon CloudWatch
- AWS CloudTrail
- AWS IAM
Business Benefits:
- 60-70% faster real-time vehicle visibility – Faster operational decisions driven by real-time telemetry and analytics.
- 50-60% faster telemetry processing – Near real-time analytics through Kinesis-based streaming and event-driven Lambda processing.
- 45-55% better battery health monitoring accuracy – Improved battery lifecycle management through SageMaker-driven predictive models.
- 40-50% improvement in predictive maintenance readiness – Reduced unplanned downtime through continuous monitoring and ML-driven maintenance insights.
- 70-80% increase in analytics scalability – A future-ready, cloud-native infrastructure that scales with fleet growth.
- 50-60% better operational monitoring visibility – Proactive issue resolution through end-to-end CloudWatch observability.
Is fleet data arriving too late to guide maintenance?
Connect vehicle telemetry with predictive insights and controlled MLOps on AWS.
Key Takeaways:
- Secure telemetry – X.509 authentication over MQTT protects the telemetry connection from the start.
- Timely operational signals – Kinesis and Lambda turn high-frequency vehicle data into timely operational signals instead of delayed batch reports.
- Controlled model releases – Staged releases and human approval help manage deployment risk as battery-health and maintenance models evolve.
Frequently Asked Questions
As its connected fleet grew, the startup struggled to process high-frequency telemetry and turn it into timely operational insight. It also needed battery-health monitoring and predictive maintenance capabilities to address service costs, unplanned downtime and uncertainty around battery degradation.
Connected vehicles needed authenticated access before sending operational data to the platform. AWS IoT Core received telemetry over secure MQTT connections, using X.509 device certificates and fine-grained IoT policies to control access.
Battery degradation and maintenance planning required insight beyond raw telemetry. Amazon SageMaker supported model development, training, deployment and retraining for battery state-of-health prediction, range estimation, energy consumption optimisation and predictive maintenance.
As models evolved, the team needed a repeatable path from training to production. SageMaker Pipelines handled preprocessing, training and evaluation, while CodePipeline and CloudFormation supported staged deployment with manual approval before production release.
Vehicle and operational data required controls across ingestion, storage and access. The platform used X.509 device authentication, encryption in transit and at rest, and least-privilege IAM access, with Amazon CloudWatch supporting operational monitoring.
