Peaksters: using AI to keep the next generation in sport
Roshini (Rosh) Bains is the Deputy Editor of Startups Magazine.…
Young people are increasingly less inclined to take up sport, according to research. A UK study of 450 14–18-year-olds, conducted as part of the ICOACHKIDS+ project involving Leeds Beckett University and European partners, titled Enhancing Youth Sport Participation in the EU: Rates, Motivations, Barriers and Ways Forward, found that 52% had stopped participating in organised sport.
Factors contributing to the dropout rates included performance-related stress, prioritising studies, a lack of self-confidence, and a lack of encouragement or enjoyment.
A growing, well-documented consequence of this is that children are becoming less healthy and less active, with 2.1 million children and young people in England being classed as less active in 2024–2025.
Vladimir Isakov noticed this problem and was prompted to launch a platform to help fix it.
Peaksters is an AI-powered personal sports development platform for young athletes, with the aim of helping young people improve their sporting performance, develop their fitness and skills, and build confidence.
Isakov, the startup’s Co-Founder and Chief Business Development Officer, sat down with Startups Magazine to provide more insight into how the platform works and the potential it offers.
What led to the creation of Peaksters, and what did you identify as the main gaps in how children develop fundamental movement skills through sport?
For more than 15 years I ran a network of five fitness clubs. Three of them had a children’s club inside, with sport and coaching sessions for children up to the age of 14. One pattern repeated itself. Coaches set homework: a few exercises for balance, core strength, or agility. Children almost never did it. There was no structure at home, no motivation, and nobody checking whether the movement was right.
The coaches were not the problem. In a group of 12 to 15 children, a coach has a session plan, a sport to teach, and very little time for one particular child. So the session goes to sport skills: passing, serving, skating. Strength, balance, coordination, and control of one’s own body get whatever time is left, which is usually not much.
The result shows up a few years later. A study published in Poland in 2025 found that roughly three in four children aged 10 to 14 in sports-focused school classes lacked sufficient basic movement skills. These are children who train regularly. The foundation under the sport skill is thin.
My Co-Founder Lira Galieva saw the same thing from the other side, as a mother of three children in sport. Her daughter is a Slovenian rowing champion and trains every day, yet even at that level basic physical preparation was a struggle. Between us we had the operator’s view and the parent’s view of one problem: the work between club sessions does not happen. Peaksters started as an attempt to make that homework actually get done.
How does Peaksters use AI-assisted coaching and computer vision to assess children’s movements and provide feedback? What can the technology identify that traditional coaching might miss?
The phone camera and an existing body tracker do the first part. They locate 33 points on the body: shoulders, hips, knees, ankles. That technology is now widely available. What matters is the judgement on top of it. Is this a correct repetition of this exercise for a child, and if not, what exactly should the child change?
For every exercise our head coach defines the few conditions that make it worth doing. In a single-leg toe touch, for example, the free leg stays off the floor and the arms straighten at the top. Our engineers turn those conditions into explicit rules. By 1 October we had more than 500 such rules across 74 exercises, each tied to a specific mistake and a specific cue the child hears, such as ‘lift your leg off the floor’. The programme has since grown to more than 90 exercises in 48 structured workouts. If the camera cannot see the body clearly, the repetition is left unscored rather than guessed.
What can it catch that a coach in a group might miss? Mostly the quiet errors. A child who rushes, swings instead of controlling the movement, or puts the foot down for a moment. A coach with 15 children sees some of these. The camera sees every repetition, every session, including the ones done at home when no coach is there. It does not read mood, fatigue, or confidence the way a good coach does. That is still the coach’s job.
What evidence is there that digital coaching can improve children’s strength, balance and coordination, and potentially help reduce the risk of sports injuries?
It’s a fair question, and I would separate what the wider research says from what we can show ourselves.
In our app, we use gamification and feedback as the key digital elements that help boost progress in completing exercises.
According to Ofcom, children aged 8 to 14 spend the equivalent of about one day per week online, and 72% of them play online games.
The motivational value of games is well described by self-determination theory. Games that satisfy the needs for competence, autonomy, and relatedness foster engagement and enjoyment. This is significant for physical training, as the benefits of strength and balance programs depend on regular and long-term adherence to a training regimen. Therefore, game design can be an effective tool for ensuring commitment to training, provided that it encourages proper movements rather than simply physical activity or screen time.
The strongest argument in favor of AI in motor skills coaching is not its novelty, but feedback, which is one of the most extensively studied variables in the field of motor learning.
A meta-analysis of sequential trials, including 15 studies (853 participants) in the field of physical education, showed that feedback improved the acquisition of motor skills compared to the absence of feedback.
The link between training load, physical foundations and injury is well documented. A study in the American Journal of Sports Medicine found that young athletes who spent more hours a week in organised sport than their age in years had about twice the odds of a serious overuse injury. Strength, balance and control of the body help reduce that risk. The Lauersen meta-analysis shows a clear gradient: strength training reduced sports injuries to less than one third, proprioceptive training almost halved them, and stretching had no measurable effect.
On digital coaching specifically, the evidence is still young, but what we can say is that by 20 October, 528 children had trained with Peaksters, active children were averaging 2.9 workouts a week, and up to 89% of those who finish a first workout start a second. We have recorded more than 300,000 child movement events at home with feedback. That is engagement. It is not yet proof of better balance or fewer injuries, and I would not present it that way.
The next step is to measure it properly. Each rule is now being tested against coach-labelled recordings of deliberate mistakes. We have also reached an agreement with a Finnish football club, whose coaches will use Peaksters in the training of their young players. Together with the club, we will work out how to measure each child’s progress. We expect to see early signs. We will report what we actually find.
How can technology make structured physical training more accessible to children outside organised sports clubs, and what role should parents and qualified coaches play?
Look at how it works for a typical family today. A child plays football twice a week, and the coach suggests extra work on balance and strength. If the parents have money, they book one-to-one sessions. If not, the child gets a printed sheet or a video link, and in my clubs most children stopped doing it within a few weeks. It depends on many factors, not only the child. But the pattern was consistent.
Technology changes the cost side of that choice. Our sessions take about 15 minutes, need no equipment and run on a phone placed on the living-room floor. The child gets a demonstration, feedback on each repetition and a reason to come back the next day. For families who will never pay for a personal coach, that is the difference between some structured practice and none.
The adults still matter. Parents set up the account and the space and, where needed, watch how their child does the session and whether they are comfortable with it. They also read the reports, which show the child’s progress and their motivation. They do not need to be coaches, because the AI coach in the app takes on that role.
Qualified coaches gain the most. In our pilot with a Finnish football club, coaches will follow each child’s progress between team sessions and adjust their own training plans in response. AI will not replace good coaches. It makes them more effective and can make sure young athletes actually do their physical preparation homework.
There is a wider point. Sport England’s latest figures show that just under half of children aged 5 to 16 in England meet its activity measure. Clubs alone cannot close that gap in the time they have.
How do you ensure AI tools designed for children are safe and age-appropriate, and avoid putting too much emphasis on performance metrics at the expense of physical development and wellbeing?
The first rule is that an adult standard is not a child standard. When we take an exercise set up to adult rules and apply it to children, the system finds a very different number of errors. Children have different body proportions and often move through different joint angles, so a movement that is correct for a nine-year-old can look wrong by adult measures.
We saw this in backward shoulder rolls. A threshold set on our coach, who is 31, rejected 28 of 43 repetitions by a nine-year-old that the coach accepted as correct. We withdrew it and now measure each child against their own starting position. A system that corrects a child for being a child teaches them to stop.
Safety is built in at several levels. Every exercise and cue is reviewed by a qualified strength and conditioning coach, and children are told to stop if they feel pain or dizziness. Routine camera analysis runs on the phone: raw video of the child is not recorded or uploaded, and research recordings need separate parental permission. Accounts are parent-led, there is no messaging between users and we work to the ICO Children’s Code. We make no diagnosis, treatment, or injury-prevention claims.
On metrics, I saw in my clubs what comparison does. A child in a group of 15 who notices she is clumsier than the others often leaves, however kind the coach is. So our rewards are tied to completed practice and the child’s own next milestone. The risk can run the other way too. Our sessions include game elements and often feel like play, and some children ask for four or five workouts a day. Rewards should encourage effort, not endless repetition, and finding that balance is work we are actively doing.
For more startup news, check out the other articles on the website, and subscribe to the magazine for free. Listen to The Cereal Entrepreneur podcast for more interviews with entrepreneurs and big-hitters in the startup ecosystem.




