Your smartphone as a biomechanical lab? Is that possible, given that AI is redefining fitness through real-time coaching and computer vision? Yes, says Winston Zin, CEO and Co-Founder of Impakt, to EFY’s Akanksha Sondhi Gaur, narrating how technology, vision, and scale come together.
Q. Can you share key company details and the vision behind the name Impakt?
A. Founded in 2020, Impakt operates as a Delaware C-Corp with a team of around 13 employees. The core founders are Winston Zin (CEO) and Hakim Bawa (CTO), along with early contributors from the initial phase. The name reflects our mission to create a meaningful impact on global health. We aim to help people live longer, healthier lives and improve how individuals approach fitness and well-being through accessible, technology-driven solutions. At the same time, we empower health-focused businesses, trainers, clinics, gyms, and educators with free tools to track, understand, and support clients at scale, even beyond in-person sessions, enabling a more continuous and effective model of care.
Q. What inspired Impakt, and what problem does it solve?
A. I come from a Silicon Valley background where innovation is part of the culture. I studied Electrical Engineering and Computer Science at UC Berkeley and wrote my master’s thesis on innovation. I have built two startups, Aeotec (now a German multinational) and Fantem (acquired by a Chinese public company), often as the first engineer, focused on combining existing ideas to create value. But Impakt is personal. I lost my sister at 36 and committed to building something in health. We address accessibility in fitness because most people lack access to affordable coaching. Our goal is to deliver intelligent training through smartphones to help people live healthier, longer lives.
Q. What is core innovation, and how does it work?
A. We are turning your smartphone into a real-time biomechanical lab. It is like an AI (artificial intelligence)-powered personal coach. The innovation lies in compressing computer vision and pose estimation into lightweight, mobile-first models that run on-device, eliminating backend latency and enabling instant feedback. The platform combines tracking across yoga, HIIT (high-intensity interval training), and callisthenics with nutrition intelligence to build a personalised dataset for each user. This enables continuous learning and optimisation towards goals such as weight loss, muscle gain, mobility, and longevity. For users, it feels like a real coach, guiding and adapting in real time while the complexity remains abstracted.
Q. How do you blend tech and business innovation to scale the platform?
A. The app integrates technological and business innovation. Our AI vision system enables live motion tracking on smartphones, while a business layer built around UGC (user-generated content) drives growth. Workouts are automatically turned into shareable highlight reels, removing content friction and turning users into creators. AI agents assist distribution across platforms. Instead of traditional affiliate models, we use a ‘stealth affiliate’ approach where users share their fitness journey and earn rewards, aligning motivation with monetisation and driving organic growth through K-factor dynamics.
Q. How does your motion analysis pipeline convert real-time video into actionable coaching?
A. The system captures video, applies pose estimation to extract skeletal key points in inferred 3D space, and maps them to structured models. A multi-state architecture breaks exercises into phases, for example, push-ups into ‘up’ and ‘down’ states, each with confidence scoring and debounce logic to ensure accuracy. Beyond motion detection, it performs biomechanical analysis to assess posture, joint alignment, and movement quality. For instance, it detects flared elbows in push-ups and provides real-time corrective feedback. Complex exercises like burpees are tracked through sequential states, ensuring complete and correct repetitions, creating a seamless coaching experience.
Q. What proprietary intelligence enables real-time corrective coaching, and why is it difficult to replicate?
A. The differentiation lies in our proprietary data and experience. We have built extensive datasets capturing real-world exercise behaviour, including errors, patterns, and user expectations from AI coaching. This data is refined for variability and used to generate lightweight on-device models. The challenge is not just detecting movement but aligning feedback with human expectations. Issues like incorrect rep counting or missing feedback can break trust. Our depth of behavioural understanding and dataset quality make replication difficult.
Q. What does your AI stack look like, and how do you manage latency and optimisation?
A. Our AI stack includes pose estimation models, multi-state motion tracking, and proprietary decision engines. We moved from approaches such as KNN (k-nearest neighbours) to a hybrid architecture that balances accuracy, latency, and efficiency. Heavy processes such as training, optimisation (including quantisation, pruning, and distillation), and handling edge cases are handled by backend toolsets, generating lightweight models for on-device inference. This ensures near real-time responsiveness with minimal battery and thermal impact, while maintaining compatibility across devices without specialised hardware.
Q. How does it incorporate longevity insights into user guidance?
A. Impakt integrates longevity insights by linking fitness, nutrition, and behavioural data to long-term health outcomes. It helps users understand how daily choices affect well-being, contextualising actions in terms of longevity, cardiovascular health, and disease prevention. By combining live feedback with cumulative health intelligence, it encourages informed decisions beyond immediate fitness goals.
Q. How would you describe it as a platform, and what architectural decisions support this positioning?
A. Impakt is a comprehensive platform combining AI coaching, motion analysis, content creation, and distribution. Architecturally, it separates heavy backend processes such as training and optimisation from lightweight on-device inference, ensuring responsiveness across devices. On the business side, it integrates UGC, as I said earlier, and automated sharing, enabling users to act as both consumers and contributors, strengthening engagement and growth.
Q. How do device limitations affect your system, and how do you ensure compatibility across smartphones?
A. Device diversity is central to our design. We avoid specialised hardware such as depth sensors and instead rely on vision-based approaches that work with standard cameras. Heavy computation is handled in the backend, while optimised models run on the device. This ensures efficiency across mid-range and high-end smartphones while managing battery, thermal, and latency constraints for consistent performance.
Q. What is the Impakt AI Fitness app, and who is it best suited for?
A. Impakt is designed for both individuals seeking instantaneous coaching and professionals such as trainers, creators, gyms, and healthcare providers who want to deliver better programmes and stay connected with clients remotely. Using AI and a smartphone camera, it tracks workouts and nutrition, provides instant feedback, and offers continuous visibility into progress, enabling better engagement and monetisation.
The Impakt AI Fitness app is free and available on the Google Play Store and at impakt.com/download. It offers AI-powered at-home workouts, tracking, and rewards. It is best suited for beginners and home users seeking equipment-free, AI-guided training with gamification, though less suited for advanced athletes requiring highly precise tracking.
Q. How do you validate the accuracy of your system and ensure reliable performance in real-world conditions?
A. We validate by aligning outputs with biomechanical standards and metrics such as intensity, movement quality, and energy expenditure models like MET (metabolic equivalent of task). While lab systems operate in controlled settings, we prioritise real-world accuracy, accounting for lighting, camera angles, and user variability. Continuous refinement using diverse datasets ensures consistent, actionable feedback outside controlled environments.
Q. How accurate is the system compared to lab-grade solutions, and what are its limitations?
A. We do not aim to replace lab-grade systems or elite coaches but to serve users needing accessible fitness solutions. Accuracy varies with real-world conditions such as lighting and setup. While performance is high in controlled settings, home variability limits precision. We track posture, joint positions, and movement quality, but hardware and environmental constraints prevent full lab-grade accuracy. Our focus is on actionable insights for consistent improvement.
Q. How is the platform designed in terms of team structure and talent philosophy?
A. We are a lean, engineering-driven team, including a CTO, an architect, engineers, technical managers, and a designer. This enables speed and focus on product development. We hire across industry and academia, combining fresh perspectives with execution strength, and prioritise first-principles thinking and openness to cross-domain ideas.
Q. What are your global strategy, growth plans, and key markets?
A. We operate as a Delaware C-Corp in the United States (US) with a growing presence in China, focusing on both markets. In the US, we empower individuals and professionals with a free AI tool, including an initiative to onboard 500 postnatal clinics. In China, we are preparing for exposure on a major entrepreneurship television show reaching over 200 million viewers, supported by a WeChat mini app. UGC drives our growth, along with organic channels, and disciplined optimisation of return on ad spend, tailored to regional behaviour.
Q. How are partnerships and ecosystem collaborations driving your growth?
A. We integrate advertising into user-generated content, enabling users to create and share branded content while earning rewards. This offers a scalable alternative to traditional marketing. We have partnered with brands such as FijiWater and Hyatt. The platform is free, with incentives driving adoption, and we are expanding partnerships globally, including in India.
Q. What is your long-term vision and message for readers?
A. Our vision is to make high-quality health and fitness guidance accessible to everyone. By combining AI, computer vision, and behavioural insights, we are building a platform that guides workouts and supports better lifestyle decisions in real time. We see strong opportunities in India and aim to scale through partnerships. For readers, if you want a smarter and more engaging way to stay fit, Impakt is built for you.
