A product engineer at Samsara with little prior data science experience built a sophisticated video analysis tool for detecting vehicle misuse. This demonstrates how accessible, powerful platforms democratize AI and blur the lines between specialized and generalist engineering roles.
An AI-optimized routing plan was rejected by a route planner because it broke established, valuable relationships between specific drivers and customers. The insight is that pure optimization is naive; successful AI must assist human workflows and account for intangible human context.
Amid a culture of hackathons and demos, Samsara prioritizes AI projects by focusing on concrete customer problems. The ultimate filter isn't technical novelty but whether the solution creates enough operational leverage that a customer would pay for it, grounding R&D in business value.
Initially addressing fuel theft for customers in one region, Samsara applied its statistical models globally. The data revealed that fuel siphoning is a massive, widespread issue, not a niche problem, demonstrating how data analysis can uncover hidden global markets.
Standard navigation apps like Google Maps can lead commercial trucks into disastrous situations, such as colliding with low bridges. Samsara built a specialized tool that accounts for vehicle dimensions, road restrictions, and even company-specific rules to ensure safety and compliance.
Samsara built a central endpoint that abstracts away complexities of using different LLMs like OpenAI or Gemini. This gateway handles cost, security, and compliance, allowing any product engineer to quickly build and deploy AI features without specialized expertise.
Samsara's AI systems, like in-cab cameras, are built to function without connectivity for extended periods (e.g., a week). They gracefully degrade and sync when back online, a crucial feature for industries like utilities construction working in areas without roads or cell signals.
Today's routing algorithms use approximations for complex scenarios. Praveen Murugesan explains that quantum computing could provide precise, optimal solutions by processing immense variables like real-time traffic across thousands of stops and multiple vehicles, moving beyond predictive models.