How we built a smart sensor for a sport that had never had one

Two years ago, David from Ve2Max came to us with a compelling idea: put a sensor inside a padel racket and give players real data about their game. Padel is one of the fastest-growing sports in the world, and nobody had done this yet. We took on the challenge and built something we're genuinely proud of.

The hardware constraints were tight from day one. The sensor has to fit inside the butt cap of a racket, survive hundreds of hard shots per session, and run long enough on a battery that players never have to think about it. We chose a Nordic nRF52840 as the core chip and paired it with a 6-axis motion sensor sampling at 230 times per second. The firmware runs on Zephyr RTOS, which we know well. Getting the state machine right across factory mode, shipping mode, and active mode, each with its own power behaviour, took careful work, but it ended up stable and power-efficient.

Bluetooth in a real sports environment is a different beast than on a lab bench. A sensor swinging at the end of a racket, a phone in someone's pocket, a busy court: it's a demanding setup. Our goal from the start was ambitious: stream full raw data live over Bluetooth so we could run advanced algorithms in the cloud rather than being constrained by what fits on the device. We tackled connection reliability systematically, adding detailed logging, integrating crash reporting via Sentry, and iterating on our reconnection logic until we had a stable experience across both Android and iOS, platforms that behave differently enough to demand platform-specific attention.

“

Building a smart padel racket with sensors, an app, and AI feedback sounds straightforward, until you actually try to do it. Insyght took care of everything: the electronics, integrating it into the racket, the firmware, and the app. One team and full ownership. And honestly, just great people to work with.

Manuel Benavente Roldán

Director at PadelPlay

Rather than compromising on our live-streaming vision, we built a resilient fallback architecture to support it. When a connection drops mid-session, a three-layer system kicks in: RAM buffering for short interruptions, flash storage for longer ones, and a full replay to the cloud when the phone reconnects. No data loss, no gaps. This keeps the raw data pipeline intact and our cloud-side algorithms running on complete sessions.

Shot classification is one of the most technically interesting parts of the product. Teaching an algorithm to tell a lob from a flat drive from a slice, out of raw IMU numbers, is hard. We calibrated with the ve2 team using sessions with professional players, and accuracy keeps improving as real match data comes in. Player feedback feeds straight back into the model.

The team is small and international, with developers in the Netherlands and Brazil working closely with the PadelPlay and Ve2Max teams on product and go-to-market. Gen2 hardware is already prototyped, with a new sensor chip that reads more accurately. There is a lot more to build.