Sibros Technologies Inc.
Production predictive ML platform on Amazon SageMaker for connected-vehicle telemetry at 75 billion data points per day — enabling a new AI/ML line of business for Sibros's international-automaker customers.
The customer
Sibros is a North-America-headquartered ISV operating in the connected-vehicle and over-the-air (OTA) software-update space. Their unified OTA-updater and telemetry platform sells to automotive OEMs and fleet operators worldwide, ingesting ~75 billion telemetry data points per day from fleets ranging from 20,000 to 400,000 vehicles each.
The challenge
Sibros's commercial model and AWS infrastructure cost are both directly driven by how much telemetry each vehicle logs — which is in turn driven by customer-defined logging rules. A given fleet configuration can produce anywhere from ~13 billion to ~40 billion data points per day. Before engaging VeUP, Sibros had no model-driven way to forecast fleet data volume before rollout, creating mispriced rollouts, AWS-cost variance, and customer trust risk on cost estimates.
Internal attempts on local Python (XGBoost / Random Forest on engineers' laptops) and on GCP BigQuery ML capped at the 70% accuracy bar. Customer-acknowledged on the 2025-10-29 project kickoff (Jiwan Senapati, Sibros, Sr. Engineering Manager): “we tried to run our data through all these models, but the 70 percent mark is hitting.”
The solution
VeUP delivered a production Amazon SageMaker training and inference platform inside Sibros's AWS account — anchored on SageMaker for managed training, model registry, and real-time inference; Amazon S3 + Apache Iceberg + AWS Glue Data Catalog + Amazon Athena as the data-lake / feature layer; AWS Step Functions + Amazon EventBridge for orchestration; AWS IAM Identity Center federation; AWS KMS encryption; Amazon CloudWatch observability. Strict GroupKFold on rule_id with 30 seeds enforces leave-rules-out generalization to continuously-added new logging rules.
The platform shipped at 74% production accuracy on the 75B data-points/day workload, exceeding the >70% contractual acceptance bar. A v2 modelling effort lifted production R² from 54% to 80%+ under the same cross-validation discipline. Sibros's engineering team operates the production pipeline independently post-handover.
Production outcomes — customer-business KPIs
| KPI | Result |
|---|---|
| Production prediction accuracy on Amazon SageMaker | 74% production accuracyon the 75B data-points/day workload vs >70% contractual minimum. R² 0.8211 on held-out test set; 88.9% of predictions within 2× actuals; leave-rules-out R² 0.73; overfitting gap −0.06. |
| New AI/ML telemetry line of business | Production SageMaker platform enables Sibros to offer predictive AI/ML telemetry products to international automaker OEMs as a new line of business that did not exist before this engagement. A v2 modelling effort lifted production R² from 54% → 80%+, expanding the LOB's accuracy headroom. |
| Customer-team independent operation | Sibros's engineering team (Jiwan Senapati, Sr. Engineering Manager; Murrilo, engineering) operates the production Amazon SageMaker pipeline independently. Knowledge-transfer sessions complete. |
Commercial signal & AWS Partner Funding leverage
Two production-extending SOWs signed with $40K aggregate AWS Partner Funding leverage— SOW #1 (Volume Prediction Model, Closed-Won 2025-09-24) at $20K T&M with a $20K AWS Partner funding offset, and SOW #2 (extension, signed 2026-04-09) at $30K T&M with a $20K AWS Partner funding offset. The customer chose to extend production engagement after production deliverable acceptance at 74% accuracy.
In the customer's words
“We tried to run our data through all these models, but the 70 percent mark is hitting. I think we need to figure out which parameters to consider and which not to consider. So that's where I think we will need a little help.”
Amazon SageMaker (training + real-time endpoint) · Amazon S3 · Apache Iceberg · AWS Glue Data Catalog · Amazon Athena · AWS Step Functions · Amazon EventBridge · AWS IAM Identity Center · AWS KMS · Amazon CloudWatch · AWS Cost Explorer.