SharpEyeWinter 2027

Train a computer vision model on real tennis matches

SharpEye's computer vision model watches match video the way a coach does: it reads every shot, who hit it and how, and turns the match into a performance report. This January a small group of students will join the team that builds it, working with one another and directly with SharpEye's co-founders.

You can join as an intern (a defined project with the team, a measured result and a reference from a co-founder) or take on a research project (an open question you investigate and write up, with a faculty advisor). Tell us which you prefer below.

DatesJanuary 4 – 27, 2027
TimeAbout 100 hours
WhereRemote
PlacesUp to five

1 · Rank the roles

Give your first choice a 1, your second a 2, and so on. Rank as many as interest you and leave the rest as "Not for me".

Shot-by-shot detection Core

The model has to name every shot in a match: the serve (flat, slice or kick), forehand or backhand, topspin or slice, volleys, overheads, lobs and drop shots, and who hit it. You build the examples it learns from, shot by shot, then use them to make it better: test it against real matches, find which shots it confuses, and measure each improvement.

Good fit: computer science, statistics, data science, anyone curious about machine learning. Tennis experience helps but isn't required.

The SharpEye stick Core

Players film their matches on the SharpEye stick. It hooks onto the fence behind the court, holds the phone at the best angle for our software, and has built-in battery charging. We have a working prototype. You would own it from here: improvements to the fence hook, phone mount and battery; materials, cost, design for manufacturing and choosing a manufacturer; packaging, shipping and getting it to players and teams.

Good fit: mechanical engineering, product design, physical prototyping.

Where the ball bounces Additional

With one camera, the model has to work out where the ball met the court without a second view to judge depth. You would work on recovering the bounce point from the ball's flight.

Good fit: physics, especially trajectory modelling, projectile motion with drag and spin, or state estimation such as Kalman filtering.

Reading the players Additional

Players call the score before a point and call balls "out". Some say "nice shot", celebrate or vent. Body language tells the same story, from a fist pump to a dropped head. These sounds and gestures tell the model when a point ended, who won it and whether its score is right.

Good fit: machine learning, language or speech processing, linguistics.

Economics of match analytics Additional

Who pays for match analytics, and how much? Players, parents, coaches and college programmes each value it differently. You would study the market: what each group would pay, how pricing by match or by shot compares with the alternatives, and what that means for how SharpEye should be sold.

Good fit: economics, business, statistics.

2 · Internship or research project

3 · About you

4 · Résumé

PDF or Word, up to 8 MB.