Reducing last-mile delivery delays by 25% for a Southern African fleet with 25,000+ annotated GPS, video, and document datasets.

Client Overview

A logistics firm specializing in cross-border trucking in Southern Africa required AI-driven route optimization to combat delays from unpredictable road conditions and customs. They needed annotated geospatial and sensor data to train models for dynamic rerouting.

Project Scope

DataLens processed 25,000+ GPS logs, dashcam videos, and shipment documents. Services covered:

• •Video frame annotation for road hazards
• •Trajectory labeling for traffic patterns
• •Document extraction for customs compliance

LLM fine-tuning enabled natural language querying of logistics data for predictive insights.

Challenges

• •Real-time data streams with incomplete sensor coverage.
• •Cross-border data sovereignty issues.
• •Handling extreme weather variability in datasets.

DataLens Solution

Employing the AI Studio's video annotation tools, DataLens integrated temporal labeling with geospatial overlays. Fine-tuning involved dataset curation for scenario-based prompts, like "reroute for flood risks," ensuring models adapted to regional logistics jargon.

Results

• •Cut delivery delays by 25% through optimized routes.
• •Achieved 96% labeling consistency and 88% model accuracy in simulations.
• •Project wrapped in 9 weeks, scaling to 5,000+ daily shipments with enhanced reliability.