Meet Our Customer

Fortune 100 Agriculture and Industrial Equipment Manufacturer

Industry

Agriculture Technology

Application

Autonomous Tractor

Results

24/7

real time coverage

21.08

median response time (in seconds)

1,280

engagements per hour

Overview

Our customer is a Fortune 100 world leader in providing advanced products, technology and services for customers in agriculture and other industries — those who cultivate, harvest, transform, enrich and build upon the land to meet the world's increasing need for food, fuel, shelter and infrastructure.

As the global population grows to 10 billion by 2050, global food demand is projected to increase by 50%. Pressures from climate change, labor shortages, limited arable land, and elsewhere are only impacting the ability of farmers to meet this growing need.

In keeping with its storied legacy of innovation to meet the moment, our customer sought to introduce a revolutionary, fully-autonomous tractor. This first-of-its-kind product would enable farmers to instantly amplify productivity and unlock hours of precious time that could then be reallocated to other important activities.

The Challenge: Autonomous Tractors & The Edge Case Problem

Our customer’s autonomous tractors feature multiple arrays of stereo cameras and other sensors that continuously feed data to a deep neural network that classifies obstacles and determines how best to respond to them.

These AI models are trained to anticipate a variety of circumstances, but the unstructured environment of a farm field is inherently unpredictable, making it impossible to train for the longtail of all possible scenarios. To be viable, autonomous tractors must be capable of consistently and confidently understanding their surroundings, detecting obstacles, and knowing how to safely navigate them.

Edge cases associated with harsh environmental conditions (e.g., dust, rain, snow, glaring sun), coupled with seasonal variations in the landscape itself make this challenging for even the best autonomous systems. Though AI development methodologies are continuously improving, this challenge persists, and with safety and reliability at stake, it represents a critical barrier between being able to launch today, or not.

“You’ve got sunny days, big trees, small trees, light soil, dark soil, ditches, cows, an ice cream truck that just pulled in. These things are incredibly hard to think about and train for"

Fortune 100 OEM executive

Without the thoughtful use of SparkAI in their production AI workflow, these hard-to-predict edge cases would have jeopardized the success of our customer’s autonomous tractor program in three key ways:

Safety

With nearly two centuries of brand reputation at stake, the requirement for safety was paramount and immutable. Introducing a product that failed to recognize an obstacle and causing bodily harm or material damage was unthinkable. Bolstering the safety case by ensuring edge cases would be handled confidently was critical to success.

Reliability

One way to ensure safety is by building an AI system that is overly cautious. Doing so, however, means sacrificing reliability. Like any expensive piece of farming equipment, uptime and asset utilization are carefully tracked by the farmer. An autonomous tractor that stopped or called for help too often would not be viable.

Cost

The need for a solution like SparkAI was clear, but enabling this functionality is complex and operationally intensive. To a company specialized in building robots and perfecting AI models, building internally would have been distracting, costly, and would have diverted precious engineering resources to an important but entirely non-core endeavor. Building the technology initially is one thing, operating continuously is another. As more autonomous tractors were deployed, the overhead of managing, training, and scaling operations would have been untenable.

The Solution: Solving Edge Cases for the World’s Most Advanced Autonomous Tractor

Here’s how it works: in moments of low-confidence, our customer’s autonomous tractors automatically call SparkAI’s service, passing imagery and other metadata via REST API. SparkAI’s objective is to resolve difficult-to-discern details about the scene to support a real-time decision.

We do this by combining two key components in real-time: (1) cognitive input from multiple human mission specialists trained for the use case, and (2) results from our own proprietary software-based decision systems. SparkAI returns this resolution directly to the vehicles. The vehicles then combine this resolution with their pre-existing knowledge of the world to decide on a safe and confident action.

SparkAI’s role in the workflow is not to take over control and remotely pilot the vehicles – our customer’s autonomous tractors are intelligent enough. Instead, what SparkAI delivers are the contextual cues that autonomous systems are sometimes missing in order to make confident decisions.

The entire round-trip process takes seconds, and happens thousands of times per day, 24/7. Demand fluctuates unpredictably, and even through moments of high activity (over 1,200 simultaneous API calls per hour), SparkAI’s median response time is 21.08 seconds — exceeding our customer’s expectations and business requirements.

Although these moments of uncertainty are rare, solving them with high-integrity enables the autonomous tractor to operate safely, reliably, and uninterrupted in the most challenging conditions at all hours of the day.

In addition to a real-time decision that supports the robot’s live action in the field, a secondary benefit to our customer is that they get production-grade ground truth data to continually improve their models. This accelerates their ML training flywheel, fast tracking a process that helps their data science team learn more about the scenarios in which their models need attention or improvement.

“Before our system is mature enough to be perfect, we can now ship it so it has the opportunity to gain that experience. Data from SparkAI then goes into our training set, and we never make that mistake again. No other farmer has to encounter it, either. The use of your machine improves the performance of all machines.”

Fortune 100 OEM executive

The Results: Resolving Edge Cases to Move Autonomous Farming Forward

At SparkAI, we believe in the promise of automation to elevate humanity. We are on a mission to help the world’s most innovative companies push the boundaries of what can be enabled with AI today.

This Fortune 100’s partnership with SparkAI accelerated their introduction of autonomous tractors, allowing farmers to capture the full value from automation today, and do more than they ever could.

“SparkAI is helping us take technology from a noun to a verb — to an activity, not a tool. The feeling of seeing a tractor do a job for a person without anybody doing anything is totally magical”

Fortune 100 OEM Executive

Our customer’s product announcement was a sensation. The company’s stock price rose sharply and the company garnered a broad base of recognition from agriculture and AI industry experts as an epicenter of automation innovation.