From operational telemetry to conversational insights: Colven's AI-powered fleet dashboard

About the client

Colven is an Argentine company with more than 50 years of experience in fleet technology, with subsidiaries in Argentina, Brazil, Mexico, Spain, the United States and Italy. It manufactures monitoring devices (VIGIA, VIESA, VITRAN, among others) and is the creator of GESTYA, its own satellite tracking and fleet data management platform, a leader in telemetry for the logistics, agriculture, mining and heavy machinery industries in Argentina, and currently expanding in Brazil with a significant presence in the agricultural sector.

Challenges

Colven already had GESTYA, its own satellite tracking platform, capable of recording key fleet events such as position, speed, ignition events and alerts, among others. Going deeper into that information, however, depended on manual processes and on reports that required time and technical expertise. Colven wanted to innovate by evolving GESTYA into a more analytical, agile and scalable platform, one able to turn fleet data into higher-value information for decision making, expanding its capabilities, optimizing operations and unlocking new growth opportunities.

At the same time, any solution had to prove that it could handle the real volume of the business and grow beyond it without major development work. In other words, the project was not looking for a simple proof of concept, but rather to validate that analytics could scale to Colven's actual operation through an MVP, before investing in a full production phase.

The Solution

Together with Colven, Teracloud designed and implemented a data pipeline, a Data Lakehouse and an analytics dashboard on Amazon QuickSuite that turns GESTYA's raw telemetry into queryable operational information. The solution ingests more than 20 million rows of data from active vehicles every day, normalizes and consolidates them into a Data Lakehouse on Amazon S3, and exposes them through a dashboard with three views (Fleet, Driver, Additional Info). The dashboard also integrates an agent powered by generative AI that answers questions about the data on the control panel, allowing each client to interact freely with their own operational data in a conversational way, which speeds up and simplifies data analysis, report creation and decision making. All of this runs with row-level security by company, so that Colven can offer differentiated analytics to each of its clients.

Project Objectives

The main objective of the project was to turn GESTYA's fleet telemetry into a robust and scalable analytics layer within AWS, one that would allow Colven to offer its clients centralized and accessible operational visibility without depending on technical expertise.

Specific objectives included:

Consolidating and modeling fleet telemetry in a reliable Data Lakehouse on AWS, sustaining a daily ingestion of more than 20 million rows of data.
Enabling centralized operational visibility of key fleet KPIs (distance traveled, speeding events, night-time usage, idling time, among others) across the entire active vehicle base, in a single dashboard.
Integrating a generative AI agent into the dashboard so that KPIs can be queried and analyzed in natural language, with no need for SQL or a dedicated analyst, accelerating analysis and decision making at that data scale.
Establishing a multi-tenant architecture (row-level security by company) capable of sustaining this volume and scaling the product to more of Colven's end clients without significant additional development.

Diagram

Key Technologies Implemented

Technical Solution and Rationale

The technologies were selected to support a serverless, transactional and self-healing pipeline at the real scale of Colven's operation:

Apache Iceberg on S3: chosen over flat Parquet because it supports native MERGE and UPDATE operations, schema evolution and concurrency through snapshots, making it possible to consolidate daily telemetry without reprocessing the entire table.
Lambda MERGE with auto-split and auto-retry: automatically splits large ingestion windows and retries on transient commit errors, ensuring the pipeline never loses a day of data because of an isolated failure.
Automatic recovery system for partial data: detects vehicles whose information was left incomplete due to API timing and reprocesses it on the following run, with no manual intervention.
Active vehicle filter: optimizes run time (from roughly 5 hours to roughly 4 hours) by discarding vehicles with no recent activity, keeping headroom within the overnight processing window agreed with Colven and GESTYA.
Amazon QuickSuite with multi-tenant row-level security (RLS): separates data access by Colven client company, making it possible to offer the dashboard and the generative AI agent as a differentiated product to each client, with no additional development.

Business Impact and Key Indicators (KPIs)

As this is a new product, Colven does not yet have adoption or usage metrics measured over time. The value demonstrated so far is technical and strategic: the solution proved that it can sustain the real volume of Colven's operation and lays the groundwork for the company to offer this capability as a differentiated service, both to its current clients and to other subsidiaries in the region, positioning Colven with a more complete and modern offering in the market.

100% data coverage across the fleet's entire active vehicle base.
Robust daily ingestion of more than 20 million rows of telemetry, validating that the architecture scales to Colven's real operation and not only to a limited pilot.
Optimized pipeline (from roughly 5 hours to roughly 4 hours of overnight processing), with operational headroom within the window agreed with GESTYA.*
Multi-tenant architecture ready to scale: Colven can offer differentiated analytics and generative AI per client with no additional development, opening the door to a new commercially viable service for its clients and regional subsidiaries.

*Note: the overnight processing window reflects an operational agreement, not a technical constraint of the architecture, which is designed to evolve toward near real-time ingestion if the business requires it.

Conclusion

Colven transformed GESTYA from an operational data source into a strategic asset. Today it can offer its clients and regional subsidiaries a modern analytics experience, with generative AI built in, that its competitors do not have. This is not just a product improvement, it is a new line of value that Colven can capitalize on commercially, setting itself apart in a market where fleet management still relies on manual reports and spreadsheets. With the technical foundation already validated at the real scale of its operation, the path to the next phase, more clients, more features and greater automation, is clear.

Get in touch

Take your business to the next level.