Why We Left SpaceX to Solve the Biggest Challenge in Hardware Engineering

The Hidden Bottleneck in Engineering’s Future

In December 2019, Boeing's Starliner was poised to dock with the International Space Station—a critical milestone in NASA's Commercial Crew Program. Instead, a software error sent the uncrewed capsule into the wrong orbit, forcing an early mission termination. Boeing later revealed that the failure cost a staggering $600 million.

For those of us who spent our careers at SpaceX designing reusable rockets and human-rated spacecraft, this was a wake-up call. One of the most expensive failures in aerospace history wasn’t caused by exotic hardware—it was a software issue. Worse, it was a preventable one.

This isn’t an isolated case. From the ispace lunar lander crash in 2023 to the East Palestine train derailment, the failure of complex machines has become routine. But these failures don’t stem from a lack of engineering talent or ambition—they stem from an inability to see and understand machine behavior at scale.

As the complexity of machines grows, so too does the risk of catastrophic errors if we don’t fundamentally rethink and modernize the software that powers them.

This isn’t an isolated case. From the ispace lunar lander crash in 2023 to the East Palestine train derailment, the failure of complex machines has become routine. But these failures don’t stem from a lack of engineering talent or ambition—they stem from an inability to see and understand machine behavior at scale.

The Complexity Crisis: Why Machines Are Failing

We are in the midst of a major shift in machine complexity. Satellites, hypersonic vehicles, autonomous robotics—these aren’t just advanced machines, they are software-driven, data-intensive ecosystems. A single test flight of an aircraft generates terabytes of telemetry, while autonomous systems process thousands of sensor inputs in real-time. This complexity is essential to their capabilities—but it also introduces failure points that are invisible without the right tools.

Despite this transformation, the tools engineers rely on to develop, test, and operate modern machines are fundamentally broken. Legacy observability tools—built for IT infrastructure, not hardware—fail under the scale and speed of modern machine data. Engineers are forced to stitch together brittle, in-house solutions just to review basic telemetry, creating a bottleneck that slows development, increases failure rates, and drives up costs.

The result? Progress stalls. Missions fail. Costs balloon. And the engineers building the future are drowning in complexity, unable to see the very data they need to make better decisions.

Engineers building the future are drowning in complexity, unable to see the very data they need to make better decisions.

The Bottleneck of Legacy Tools

Why hasn’t observability caught up with modern hardware? Because building a system that can ingest, analyze, and act on high-velocity, high-cardinality machine data in real time is an enormous challenge. Historically, companies have had to build these tools in-house, pulling their best engineers away from core product development.

The result is a patchwork of bespoke systems—fragile, expensive, and incapable of scaling. These systems demand constant maintenance, and when a company tries to scale or adapt, they break. Engineers spend thousands of hours manually reviewing data, debugging failures, and rebuilding broken in-house telemetry tools—wasting time that could be spent pushing the boundaries of what’s possible.

This isn’t just inefficient—it directly limits progress. Every hour an engineer spends fighting with data is an hour not spent solving the engineering challenges that define the next generation of aerospace, robotics, and energy systems.

The Future is Observable: Why We Built Sift

For centuries, human progress has been defined by those willing to push the limits—explorers, engineers, and builders who saw obstacles not as barriers but as challenges to overcome. The last century saw an unprecedented acceleration in technological breakthroughs, from the Wright brothers’ first flight to the Moon landing in just 66 years. But today, we face a paradox: while our ambitions are greater than ever, the complexity of modern machines has become an anchor slowing progress.

Hardware programs have grown so cumbersome that we measure progress in decades, not years. Development cycles stretch endlessly, not due to a lack of ideas, but because the tools used to test, validate, and operate complex machines remain fundamentally broken. Engineering teams spend more time wrangling data than solving the real challenges at hand.

Modern hardware engineering faces three fundamental challenges:

• Complexity of individual machines – Advanced systems like rockets, autonomous vehicles, and satellites integrate millions of hardware and software components, requiring real-time diagnostics that legacy tools can’t handle.
• Complexity of large organizations – Engineering teams span thousands of specialists across disciplines, yet lack the collaborative infrastructure that enables software developers to collaborate on open-source projects across the world.
• Complexity of scaling fleets – As prototypes become global deployments, teams must operate exponentially growing fleets without exponentially growing headcount.

At Sift, we believe complexity shouldn’t be a bottleneck—it should be a launchpad. We built Sift to give engineers the infrastructure they need to keep pace with their ambitions. By replacing fragmented, in-house telemetry stacks with a unified observability platform, Sift transforms machine data into real-time intelligence.

This isn’t just about building better software. It’s about enabling machines without limits.

Observability Reimagined: Engineering Without Limits

For decades, engineering has been constrained by the limits of manual telemetry review and fragmented toolchains. Sift removes these barriers, providing a real-time, AI-powered observability layer for the most complex machines in the world. But Sift isn’t just another tool—it’s a paradigm shift in how complex machines are developed and operated. By unifying telemetry monitoring, storage, visualization, and review, Sift is breaking down data silos, streamlining workflows, and enabling engineers to move beyond reactive troubleshooting to proactive optimization.

Imagine a future where:

• Hardware teams can iterate as fast as software teams, eliminating costly delays in development cycles.
• Missions are no longer jeopardized by preventable software errors.
• AI-assisted root cause analysis becomes the norm, cutting debugging time from days to minutes.
• Engineers spend less time fighting data and more time solving the hardest problems in technology.

This is the future we’re building with Sift. A future where machine complexity is no longer a bottleneck—but a launchpad for the next era of human achievement.

The next generation of machines will define the future of aerospace, transportation, and energy. But without the right tools, their complexity will hold us back.

That’s why we left SpaceX to build Sift. Not just to fix telemetry, but to remove the biggest obstacle standing in the way of engineering progress itself. 

Legacy tools force engineers to react. Sift gives them the power to predict, prevent, and accelerate progress.

The Opportunity Ahead

Sift isn’t just a product—it’s a shift in how machines are built and operated. As we expand our platform, we’re abstracting away one of the hardest parts of testing, manufacturing, and operating machines, giving hardware teams the power to focus entirely on their missions.

We envision a future where engineers are empowered not just to build the next iteration, but to leap into uncharted territory. A future where the biggest challenges—space exploration, next-generation defense—are met with tools that make engineering limitless.

Sift is the foundation for that future. And we’re just getting started.