Chinese Hardware Firm Launches AI Golf Coach Using World Models for Personalized Training

Seven years ago, Jijia Technology began quietly building smart home cameras and video doorbells in Beijing. The company grew to hundreds of millions in annual revenue, millions of users across 150 countries, and a roster of serious investors—Linear Capital, Matrix Partners China, Vision Capital, and others—yet remained largely unknown outside its sector. The reason for the low profile was deliberate. While most hardware manufacturers treat artificial intelligence as a feature to polish later, Jijia's founder and CEO Li Pu had already decided that AI would be the core kernel running through every product from conception onward.

In 2024, Jijia turned toward sports hardware, choosing golf as its entry point. On September 17, the company's sports technology brand NeoPace released the G10 Pro, a launch monitor that crowdfunded for over 2 million yuan on Kickstarter. The device is notable not for measuring ball flight—that problem was solved decades ago—but for understanding the person holding the club.

Golf training hardware has long been a mature market. Traditional monitors capture ball speed, launch angle, spin rate, and trajectory with extreme precision. This data serves professional players well: they can adjust equipment and refine swing mechanics based on exact numbers. But for amateur and advanced enthusiasts, the system has a critical gap. The hardware tells you where the ball went. It does not explain why. Without understanding cause, there is no way to correct the next swing.

Jijia approached the problem from the opposite direction. During a golf swing lasting less than two seconds, the human body activates 206 bones, roughly 360 joints, and more than 600 muscles in tightly synchronized sequence. Any tiny deviation in any link amplifies into visible error when the club strikes the ball. To understand this chain of causation, Jijia built the G10 Pro on a proprietary system called TrainerIQ, a world model trained to invert 3D human movement from single-camera footage.

The technical challenge is severe. Most golfers practice alone on driving ranges or indoor bays, making multi-camera setups impossible. The device sees only one angle, with limited information. Critical details—hip rotation, center of gravity transfer, wrist release—are easily blocked or missed. Jijia solved this by building an exclusive golf sports database of hundreds of thousands of high-precision movement samples. During training, the model receives multi-angle recordings to learn standard swing mechanics. Then it learns to reconstruct full 3D movement from single-view images, using the other angles to verify accuracy. This full-domain data compensates for the information loss inherent in single-camera capture.

The result is a device that does not merely record. When a user hits a ball, G10 Pro analyzes shoulder, arm, and hip displacement in the front-view image, applies learned biomechanical rules, and deduces the complete 3D movement sequence. It judges whether the head moves, whether the waist or hand initiates the downswing, whether force sequencing is efficient. Then it converts this analysis into actionable feedback tailored to that specific person's body and habits. The system does not impose a standard answer. Instead, it simulates what the swing trajectory would look like after a suggested adjustment, showing the user what might happen to the next shot if they correct posture, optimize force application, or adjust rhythm.

Li Pu positioned G10 Pro not as a comprehensive tool but as an assisted training device. Rather than overwhelming users with every movement problem at once, it identifies which issue should be corrected most urgently and focuses feedback there. The hardware itself is merely an entry point. The real business model is subscription services—the beginning of a long-term relationship between brand and user.

Jijia's ambition extends beyond golf. The company is building a layered platform designed for reuse across multiple sports. The bottom layer is general human perception technology: skeleton calculation, force sequence deduction, single-view 3D reconstruction, and movement causal attribution. These capabilities are not bound to golf but apply to any human movement analysis. The middle layer is AI-native hardware engineering—perception frameworks, algorithm compensation for low-power devices, imaging adaptation for outdoor conditions, lightweight alignment systems. The top layer is commercial infrastructure: overseas channels, community operations, subscription models, and user retention systems developed during years in the home hardware business. Other sports products are in planning and expected to launch in early 2025.

For decades, mass sports training has moved toward standardization and efficiency. Professional hardware refreshed data accuracy. Templated error correction became mainstream. Complete training solutions flowed to enthusiasts at low cost. Training improved. Individual expression diminished. With AI that understands human movement at scale, ordinary users can now have exclusive intelligent coaching assistants. This is the optimization direction of AI-native hardware: not a smarter tool, but a device that truly understands the specific person using it.