Bringing AI to Tennis with Swupnil Sahai

Summary

Swupnil Sahai, co-founder of SwingVision (marketed at the time as “Swing Tennis”), explains the origins and capabilities of an AI-powered tennis analytics app built on computer vision technology he developed while working on Tesla’s Autopilot self-driving system. The app uses an iPhone mounted on a fence or tripod to automatically track ball placement, shot types, serve percentages, court positioning, and rally statistics — generating data previously available only to Grand Slam professionals. The product was rebranding from “Swing” to “SwingVision” at the time of recording, with investors James Blake (Harvard/ATP) and Andy Roddick (ATP) on board. The episode covers the technical architecture, use cases for junior and college players (including recruiting highlight reels), the parent-player live score streaming feature, and the future roadmap including automated match scoring and dead-time removal.

Guest Background

Swupnil Sahai grew up in the San Francisco Bay Area, played tennis starting around age 6–7 (introduced by his Indian parents), played high school tennis at Saratoga High School, and continued playing at Berkeley (where he majored in math and statistics) and at Columbia for intramural tennis. He earned a PhD in Statistics from Columbia University (2013–2017), interned at Tesla’s Autopilot self-driving team twice, and joined Tesla full-time before leaving to pursue SwingVision full-time. His co-founder is a Berkeley roommate and electrical engineer. Sahai is a USTA recreational league player. James Blake (investor and advisor) was brought in through a mutual connection — a physio who ran “Cramps Away,” a swirling mouth solution used by players including John Isner and Stan Wawrinka. Andy Roddick came on board after Sahai met him at a Houston exhibition match (alongside Jim Courier and John McEnroe) and helped Roddick recover a lost iPhone using iCloud’s remote play-sound feature.

Key Findings

1. Tesla Autopilot as the Conceptual Bridge

Sahai’s career breakthrough insight: the computer vision algorithms he was building at Tesla to track pedestrians and other vehicles on the road are structurally identical to what is needed to track a tennis ball and players on a court. The technical feasibility of bringing professional-grade ball and player tracking to a consumer phone became clear to him while working at Tesla — “someone else is going to do it if we don’t.” Apple’s simultaneous release of Core Machine Learning (making it practical to run complex ML models on iPhones) and specialized neural processing chips in the iPhone made the timing align.

2. Product Architecture: iPhone as a Multi-Camera Analytics Platform

The setup: mount an iPhone on a fence post or tripod behind the baseline, press record, play tennis. The app then automatically produces: shot type tracking (forehand, backhand, serve, volley), ball placement heat maps (where shots land on the court — cross-court forehand depth, serve placement by ad/deuce and wide/T), player positioning data (inside the service box vs. behind the baseline), serve percentage (first/second), and trends across sessions (comparing today’s stats to a 5-session rolling average). At the time of recording, automated match scoring — removing dead time and stepping through points — was in development and near release.

3. Democratizing Pro-Level Data

Sahai notes that even ATP touring professionals outside the top 50 don’t consistently have access to this kind of match data — they lack the financial resources to hire analytics staff or install court systems for every practice session. James Blake confirmed this in their pitch meeting: “even when he was training he didn’t really have access to this kind of data unless it was in a big stadium match.” SwingVision’s iPhone-based model brings Grand Slam-level shot analytics to any player with an iPhone and a $20 fence mount.

4. Recruiting Highlight Reel Generation

Players can tag favorite points within a match, filter by point type (break points won, topspin winners, serve aces, etc.), and export a custom highlight reel as a sharable video clip. Sahai described a planned feature for building multi-match recruiting reels — curating across multiple matches for the best examples of specific skills. This directly addresses a pain point Lisa Stone raised: college coaches want curated, data-annotated clips, not raw 90-minute match footage.

5. Live Score Streaming for Remote Parents and Fans

The watch app version of SwingVision already included live score streaming — parents could follow a child’s match score in real time from anywhere in the world. Sahai designed this feature specifically after watching his own parents miss high school matches. Stone identifies this as a solution for a college tennis gap: most college matches take place at facilities without any broadcast presence, and parents (and fans) learn scores only after the fact. Sahai confirms live video streaming is roadmap-planned pending user demand confirmation.

6. James Blake and Andy Roddick as Investor-Advisors

Blake participated in feature brainstorming sessions and provided athlete-perspective insights on what data would be most useful for players at various levels. Roddick (who Sahai notes was making his first tennis-specific investment) engaged substantively on product features. Both helped with introductions to USTA leadership and major US academies for partnership discussions. Their social media promotion was how Stone discovered the product. Blake and Roddick represent the bridge between SwingVision’s consumer app positioning and the professional tennis infrastructure that validates the product.

7. Trend Comparison as a Coaching Communication Tool

The stat trending feature — comparing today’s session to a 5-session rolling average — creates a neutral, objective diagnostic that both players and coaches can use without emotional subjectivity. “Today you missed a lot of forehands on break points and normally you don’t” becomes an evidence-based coaching conversation rather than a perception-based one. This mirrors the dynamic Lisa Stone described with AccuTennis two episodes earlier (ep. 2019-09-24): data as a neutral arbiter in coaching relationships.

8. Professional Player Data Gap Reveals Market Opportunity

The revelation that touring professionals below the top 50 have limited access to analytics data (because they can’t afford coaching staff or expensive court systems on road) establishes that SwingVision’s market extends beyond recreational and junior players into the professional circuit. Sahai frames this as a compelling future direction — a product that democratizes analytics not just for juniors and recreational players but for lower-ranked professionals who are currently data-blind about their own games.

Actionable Advice for Families

• Junior players should record matches and practices using SwingVision or similar apps and build a habit of reviewing shot placement data after sessions — the feedback loop of record → diagnose → practice → record is how the tool is designed to drive improvement
• Recruiting players should use SwingVision’s highlight reel feature to curate match clips by point type (not raw match video) when sending video to college coaches — annotated, filtered clips are what coaches actually want to review
• Parents who are remote from matches (at work, traveling) should use live score streaming features to stay connected without requiring physical attendance, which addresses both the FOMO problem and the over-present parent problem

INTENNSE Relevance

• SwingVision as a league analytics partner: At the time of recording, SwingVision was already named as a CoachU partnership (per the NIL/mentoring episode from 2023-04-19). The platform’s shot tracking, placement data, and automatic dead-time removal are directly applicable to INTENNSE broadcast production — auto-generated clips, shot stats, and heat maps could populate broadcast graphics in real time
• Democratizing broadcast analytics: Sahai’s core argument — professional-grade analytics should be accessible to any player at any level — maps precisely onto INTENNSE’s broadcast vision. The league can use SwingVision-style infrastructure to give every match the kind of statistical presentation currently reserved for Grand Slams, at dramatically lower cost
• Player positioning data: SwingVision’s court positioning tracking (inside service line vs. behind baseline) is directly relevant to INTENNSE’s tactical broadcast narrative — commentary about how players are positioning relative to INTENNSE’s specific format adjustments (one serve, rally scoring) would be richer with real-time positioning data
• Remote fan scoring: The live score streaming feature is a template for how INTENNSE should think about fan engagement for away audiences — not just match-specific but session-by-session across a team’s regular practice week, building the kind of continuous connection that traditional sports create through beat reporting
• Professional player data gap: Sahai’s revelation that touring professionals outside the top 50 lack match analytics validates INTENNSE’s value proposition to players at exactly the level the league is targeting — players who are good enough to compete professionally but not rich enough to have full data infrastructure support
• Andy Roddick and James Blake involvement: Both are prominent college-to-pro players (Georgia and Harvard respectively) in exactly INTENNSE’s target player demographic profile. Their enthusiasm for democratizing tennis analytics provides signal that this is the right investment direction for the league

Notable Quotes

“This is the future of our sport.” (Lisa Stone, on SwingVision)

“Even when James was training he didn’t really have access to this kind of data unless it was in a big stadium match. If he was on the practice court he also didn’t have the data — that blew me away.”

“Someone else is going to do it if we don’t do it — so that’s why we decided to go do it.”

“You’re going to see all kinds of metrics around shot placement — like are your cross court forehands going in deep? When you’re hitting serves, how often are you making serves on the ad court wide versus down the tee?”

“The timing of kind of the technology all kind of came together — Apple released Core Machine Learning, they designed new chips on the iPhone — that’s when I made that leap.”