🏇
Live · poloreport.comUSPA + HPA

Polo Platform

AI match analytics · Video pipeline · Umpire portals

AI powered sports analytics · Video engineering · Multi tenant security

AI powered match operations platform built in association with the United States Polo Association and the Hurlingham Polo Association: AI video frame analysis, team & player analytics, predictive modelling, and an MP4 to HLS video pipeline with frame accurate seeking for umpire review.

2

Independent governing bodies on one platform

4

Live portals: public, team, USPA umpire, HPA umpire

Frame

Accurate seeking for umpire incident review

The Challenge

Polo match operations span two independent governing bodies (USPA and HPA), competing teams, and umpires adjudicating live incidents: all requiring access to the same platform but with strict data isolation. Match video runs four to six chukkas at multi hour length, making progressive MP4 delivery a non starter for both bandwidth and seek behaviour. Umpires adjudicating incidents need frame accurate seeking, not “somewhere within six seconds.” And schedule integrity across clubs and seasons is unglamorous but load bearing: every downstream artifact hangs off it.

Platform Surface

poloreport.com

Public facing

Reporting and match content for fans, sponsors, and media: the public window into every tournament.

team.mysportsanalysis.com

Team portal

Team management, match review, video access, and performance data: scoped to each team's own data via row level security.

umpire.mysportsanalysis.com

USPA umpire portal

Officiating assignments, incident review with frame accurate video seeking, and schedule management for USPA credentialed umpires.

hpa-umpire.mysportsanalysis.com

HPA umpire portal

Parallel umpire deployment for the Hurlingham Polo Association: separate governing body, fully isolated data, identical tooling.

What We Built

An AI powered multi tenant sports operations platform with AI video frame analysis, team and player performance analytics, predictive match modelling, tournament scheduling, team and umpire portals, and a full MP4 to HLS video pipeline: all enforcing cross tenant data isolation through PostgreSQL row level security rather than application level WHERE clauses.

Outcome

Four live portals serving the USPA and HPA simultaneously: public reporting at poloreport.com, team management, and dedicated umpire portals for each governing body. AI video frame analysis flags potential incidents automatically, reducing umpire review time from hours to minutes. Teams access AI driven performance analytics scoped to their own data. Player pattern recognition surfaces scouting intelligence invisible to manual observation. Two governing bodies operate on a single platform with zero data leakage by design.

Quick facts

  • Platform

    Polo Report + MySportsAnalysis

  • Partners

    USPA & HPA

  • Live at

    poloreport.com · team / umpire portals

  • Domain

    Match operations · Video · Officiating

  • Category

    Sports SaaS · AI analytics · Video engineering

  • Security model

    PostgreSQL row level security

  • Stack

    Node.js · Fastify · Vue.js · PostgreSQL · RabbitMQ · Redis · ffmpeg · AWS

Build the next one

Same security model, same video pipeline: for your sports, media, or multi tenant platform.

Schedule a call

Video pipeline

MP4 in, frame accurate HLS out.

Video pipelines are where projects quietly lose months. Rendition switching, seek accuracy, audio drift, and storage cost curves only appear at real volume with real users. We built this one to survive all of them.

01

Adaptive bitrate ladder

Multi hour match MP4s transcoded via ffmpeg into 1080p / 720p / 480p / 360p renditions: viewers on club wifi and mobile both get a usable stream.

02

Keyframe aligned GOPs

Non negotiable. If keyframes don't align across renditions, the player can't switch cleanly and the viewer sees stalls at the worst possible moment. The single most common way ABR ladders are built wrong.

03

Frame accurate seeking

The entire point for umpire review. An umpire adjudicating a live incident needs to land on the frame, not within six seconds of it.

04

Async transcode queue

Multi hour source at four renditions is a long running CPU bound job that cannot sit inside a request cycle. Asynchronous job queue with compute scaled for the workload.

05

CDN delivery + access control

Object storage plus CDN with access control on protected match content. Teams see only their own matches; governing bodies see their jurisdiction.

06

HLS segmentation

Short chunks with a master playlist referencing per rendition variant playlists: progressive MP4 delivery is a non starter for multi hour polo matches.

AI powered sports analysis

AI that transforms match footage into actionable intelligence.

Machine learning models trained on match data across both governing bodies deliver capabilities that manual analysis cannot replicate: from automated video frame analysis for umpire review to predictive match modelling and player scouting intelligence.

AI · 01

AI video frame analysis

Machine learning models process match footage frame by frame: detecting player positions, horse movement patterns, stick contact, and foul incidents. AI flagged moments surface automatically in umpire review, reducing hours of manual footage scrubbing to seconds.

AI · 02

Team performance analytics

AI models analyse team level patterns across matches and seasons: offensive/defensive efficiency, chukka by chukka momentum shifts, handicap adjusted performance metrics, and tactical formation analysis. Teams see actionable data, not just raw stats.

AI · 03

Player pattern recognition

ML models trained on match data identify individual player tendencies: ride off timing, shot selection under pressure, positional discipline, and fatigue curves across chukkas. Surfaces scouting intelligence that manual observation misses.

AI · 04

AI assisted officiating

AI pre screens match footage for potential fouls, dangerous riding, and line of the ball violations: presenting umpires with AI tagged incident clips ranked by confidence, alongside the frame accurate HLS seeking for manual verification.

AI · 05

Predictive match modelling

Tournament level AI models forecast match outcomes based on team composition, recent form, playing surface, and historical head to head data: surfacing insights for commentators, analysts, and tournament organisers.

AI · 06

Automated highlight generation

AI identifies high action moments (goals, near misses, exceptional plays) and auto generates highlight packages from full match footage, with intelligent shot selection and contextual metadata.

Security architecture

Authorization enforced in the database, not the application.

The most defensible architectural decision in the platform. A missing tenant filter in one endpoint returns zero rows instead of another team's data. The developer cannot write the vulnerability by accident.

01

OAuth at the edge

Identity provider issues tokens carrying principal and claims: team membership, umpire credential, governing body, subscription tier. API Gateway validates at the edge before traffic reaches compute.

02

Claims bound to database session

Lambda extracts verified claims and binds them to the PostgreSQL session (SET LOCAL inside the transaction). Every query runs under the identity context: no ambient authority.

03

Row level security policies

RLS policies on every tenant scoped table filter rows against session context. A missing WHERE clause returns zero rows instead of another team's data. The developer cannot write the vulnerability by accident.

04

Cross tenant isolation

Two independent governing bodies (USPA and HPA), competing teams, and umpires adjudicating live incidents. A team seeing another team's video or an umpire assignment leaking across bodies is an incident with a governing body attached. RLS makes that class of failure structurally difficult.

Shared infrastructure

Same AWS spine
as the racing pipeline.

Both the racing and polo platforms run on the same infrastructure layer: API Gateway with OAuth at the edge, Lambda for compute, PostgreSQL with row level security, Redis for hot path caching, and RabbitMQ for event driven orchestration. The polo platform adds the video transcode pipeline on top.

Why RLS matters here

A team seeing another team's match video is not a bug: it's an incident with a governing body attached. An umpire assignment leaking across USPA and HPA is a governance failure. Row level security makes these failures structurally difficult rather than a matter of code review discipline.

Lessons for enterprise, brands, scaleups & VC backed founders

What a multi tenant sports platform teaches you.

01

Enforce authorization in the database

In a conventional design, a single missing WHERE clause leaks another team's data. Under row level security, that query returns zero rows. The database refuses. When the stakes involve governing bodies and live officiating, this is the only defensible architecture.

02

Video pipelines eat months silently

Rendition switching, seek accuracy, audio drift, storage cost curves: failure modes that only appear at real volume with real users. Characterising this as 'a huge task in itself' is accurate and worth stating plainly.

03

Schedule integrity earns trust

Tournament and match scheduling is unglamorous, and it's where the platform either earns trust or loses it. Every downstream artifact (video, umpire assignment, team review) hangs off schedule integrity.

04

Serverless fits spike workloads

Polo and racing traffic is violently spiky. Raceday/tournament concurrency bears no resemblance to a quiet Tuesday at 03:00. Lambda absorbs the spike and costs nothing between events.

More case studies

Same team, more categories in production.

See all case studies
P

PACE Racing

AI sports analytics · 25 year archive

AI powered thoroughbred racing analytics. Entity resolution across 3 providers, AI video frame analysis, predictive modelling, and sectional analysis from 25 Hz GPS telemetry across a 25 year archive.

AI analytics · 25 year archive
R

RYVYL

NASDAQ payments platform

Formerly GreenBox POS. Multi currency acquiring, real time settlement, treasury dashboards: engineered to a $43.5M NASDAQ IPO in 2021.

NASDAQ: RVYL · $43.5M IPO
S

The Sufferfest

Consumer SaaS · Acquired

Napkin to acquisition consumer platform. $20M+ revenue, 70+ countries, 125,000+ community: acquired by Wahoo Fitness in 2019.

Acquired · $20M+ revenue
m

mydiveo

Forbes featured · Acquired

Mobile portfolio platform for artists. $4M raise, tier 1 press (Forbes, LA Times, Billboard), acquired for $7.4M within 15 months of launch.

Forbes featured · $7.4M exit
S

SAFER

GovTech · Patented biometric

World's first biometric activated personal emergency alert system. Patented Jan 2017, beta tested with U.S. police departments.

Patented · U.S. law enforcement beta
R

REAP Pro

PropTech · Business Insider

Real estate appraisal platform replacing 70% manual data entry with OCR driven automation. Launched on Business Insider.

+20% efficiency · Hours→minutes
C

CaptureProof

AI health · Patented · $6.15M raised

AI driven visual health platform: Smart Medical Camera turning any smartphone into a clinical imaging device. 3 patents, 8+ awards including MedTech Breakthrough.

$6.15M raised · 3 patents
س

Saleem

HealthTech · 50K+ downloads · Iraq

Home healthcare marketplace for Iraq: doctor visits, nursing, physiotherapy, lab tests at your door. Arabic first, cross platform, Zain Great Idea Top Innovator.

50K+ downloads · Zain accelerated

Building a sports platform, video pipeline, or multi tenant SaaS?

Ship the platform governing bodies trust.

30 minutes with the team that built frame accurate video, row level security, and multi governing body isolation: for two of the sport's highest authorities.

Book a scoping call