Cursor Boston × AIC × Hult International — Boston Tech Week Speakers & Workshop
Full day at Hult — LSE guest lecture from Antonio Mele, lunch, 2-hour hackathon sprint, live pitches
- Date
- 2026-05-26 · 10:00 AM – 4:00 PM ET
- Venue
- Hult International (exact address after approval)Register on Luma to see the venue address after you're approved.
- Submission branch
- sports-hack-2026-submissions
How to win a Cursor credit
We have 119 Cursor credit codes for participants. They go to the top 119 ranked attendees on the website signup list — by merged PRs to the community repo, with signup time as the tiebreaker.
- 1
Use the website signup list
The list at /hackathons/sports-hack-2026/signup is the ranking source for participants. It is separate from Luma registration, which controls event admission.
- 2
Merge PRs into cursor-boston
Every merged PR moves you up the leaderboard. Browse open issues — first-time contributors should look for
good first issue. - 3
Stay in the top 119
On event day the top 119 get a Cursor credit code. Watch your rank live on the signup page — Cohort 1 applicants get a priority boost.
How to submit your project
On event day you submit by opening a PR into the sports-hack-2026-submissions branch — one folder per attendee, named after your GitHub handle, with a single meta.json describing the project. Once a maintainer merges through to main, your card appears in the grid below.
Let your agent submit for you
Copy this self-contained prompt and paste it into Cursor, Claude Code, or any agent with shell access. It asks you for the title, description, video, repo, and deploy URLs, validates them, forks this repo, writes meta.json in the right place, and opens the PR against sports-hack-2026-submissions before the 4 PM ET cutoff.
Or do it by hand
- 1
Fork the repo
Click Fork on cursor-boston so you can push without write access to the upstream repo.
- 2
Add your folder + meta.json
Create
sports-hack-2026-submissions/<your-gh-handle>/meta.jsonwith title, description, video URL, repo URL, and deployed URL. - 3
Open PR into
sports-hack-2026-submissionsSet the base branch to sports-hack-2026-submissions, not
main. - 4
Wait for the deploy
A maintainer batches merges into
sports-hack-2026-submissions, then shipsdevelop→main. Your card appears below after Vercel redeploys.
⏱ Hard deadline: 4:00 PM ET on Tuesday, May 26
Your PR must be opened before 4 PM ET (event end) to be eligible for AI scoring. One second late and your AI eval is gone — but human judges still review every merged submission, so you can still win on the judges track.
meta.json template
{
"title": "Short, specific project title",
"description": "1–3 sentences. What did you build? What problem does it solve?",
"displayName": "Your name as you want it on the page",
"videoUrl": "https://www.loom.com/share/...",
"repoUrl": "https://github.com/your-handle/your-project",
"deployedUrl": "https://your-project.vercel.app",
"tags": ["nba", "live-stats", "next.js"],
"collaborators": [
{ "displayName": "Pat Collaborator", "githubHandle": "pat-collab" }
]
}Full rules + field reference: sports-hack-2026-submissions/README.md
How winners are picked
Two independent tracks, both running off the same submissions:
AI track — 3 winners
Closes 4 PM ETAfter the 4 PM deadline, an AI agent reads every eligible meta.json + video/repo/deployed link and scores the projects 1–10 with a written rationale. Top 3 win.
Judges track — 3 winners
Open to late submissionsA panel of human judges reviews every merged submission — AI-eligible or not — and picks their top 3 based on the live presentations and the repo + deployed project.
Public submissions board
Merged submissions (31)
Cards are sorted by combined AI + judge score. Only PRs opened before 4 PM ET on May 26 are eligible for the AI track; later PRs are shown separately and remain eligible for the judges track.
0
AI-eligible
31
After 4 PM
After-deadline submissions (judges track only) (31)
Cornerman — a boxing tutor that asks the question instead of giving the answer
AI 9.0Most AI tutors are answer machines that tell athletes what to fix. Cornerman is feedback architecture: the athlete states an intent, attempts on video, predicts what the AI will say, and only then sees a question pointing at one moment with the cited still-frame as evidence. Round 2 measures whether the focus flagged in round 1 was actually addressed. Built against all six principles of the Builder Playbook for AI-enhanced learning.
athletes-performanceai-tutorfeedback-architectureboxingstreamlitopenaiAI judge: Clear #1 in the AI track. Contrarian pedagogical thesis — "AI tutors as question-asking architecture, not answer machines" — backed by a concrete 2-round structure: athlete states intent, attempts on video, predicts the AI feedback, then sees a Socratic question pointing at one still-frame as evidence; round 2 measures whether the focus from round 1 was actually addressed. Built-in evaluation design is rare for a one-day build. Sport-specific (boxing), all three URLs filled (Loom + repo + Streamlit deploy), multimodal AI use, and explicitly grounded in a named learning-design framework. Stands alone above the rest of the field.
Apex AI — Vision-Only AI Race Engineer
AI 8.5Apex AI analyzes motorsport and cycling race videos using Gemini 2.5 Flash to detect braking errors, racing line issues, and best moments with exact timestamps. Supports karting, motorcycle racing, and road cycling. No sensors or hardware needed — just video. Built with Streamlit, Groq LLaMA, and a custom interactive timeline.
kartingcyclingai-coachingcomputer-visionstreamlitgeminiAI judge: Strongest technical bet in the field — pure computer vision over race footage to surface braking errors, racing-line issues, and best moments with timestamps, no sensors required. Multi-sport coverage (karting + motorcycle + road cycling) shows genuine generality rather than a one-track demo. Named tech stack (Gemini 2.5 Flash for video + Groq LLaMA + custom interactive timeline), all three URLs filled (Vimeo + repo + Streamlit deploy). The "no hardware needed, just video" framing is a real differentiator vs sensor-based competitors. Slightly held back from 9 because the demo is a single-flow Streamlit app rather than something with sample-output galleries or a measurable accuracy claim, but the technical floor is the highest of the 8-tier.
FANCAST + StatsCast
AI 8.3by Abishek M
Talk to a live NBA game using your voice — FANCAST uses Gemini Live native audio to answer fan questions about scores, plays, and player lines in real time. StatsCast pairs it with streaming AI analyst breakdowns and animated season stats for any player or sport.
gemini-livenbavoicereal-timeaudioai-analyticsAI judge: Most contemporary AI bet in the field — voice-first interface to a live NBA game via Gemini Live native audio is genuinely novel, since most sports apps are visual-first and conversational live-game Q&A pushes the edge of what current models can do. Companion StatsCast layer (streaming analyst breakdowns + animated season stats) shows ambition without losing focus. All three URLs filled (YouTube + repo + stable Vercel deploy). Held just under the top because the meta sells the experience but hand-waves the hardest part — how live-game data integration is actually grounded — and pairing two products in one submission slightly diffuses the pitch.
CoachLens — AI Basketball Practice Planner
AI 8.2by riqthedev
CoachLens turns a coach's team identity, game issues, and notes into root-cause analysis, practice priorities, drills, and one measurable next-game target. It solves the gap between watching a game and knowing what to run at tomorrow's practice.
basketballcoachingnext.jsAI judge: Sharpest workflow design among the coaching tools: takes a coach's team identity + game issues + notes and produces root-cause analysis → practice priorities → drills → one measurable next-game target. Naming "the gap between watching a game and knowing what to run at tomorrow's practice" identifies a real coach pain point in one sentence, and the "one measurable target" output shows discipline most one-day builds skip. Sport-specific (basketball), all three URLs filled (Loom + repo + Vercel deploy). Held a touch below the top 8s because the meta doesn't call out which AI patterns or models do the analysis — the workflow is clearly LLM-backed but the leverage signal is implicit.
Celtics Green or Mean — AI Fan Cam
AI 8.0by Derek Yimoyines with Derek Yimoyines
Fans scan a QR code, snap a selfie, and pick Green (Hype) or Mean (Roast). Claude Vision scores each photo for energy, Celtics spirit, and originality, then generates playful PA-announcer roast lines. An operator dashboard queues the best fans for the jumbotron. Presentation: https://docs.google.com/presentation/d/1RMdtYTSITz0C0YzBoQAEVWxwoYdjaHzV1Q_2e8DRmfA
nbacelticslive-eventsnext.jsclaudevisionAI judge: Most production-shaped of the 8s — a two-surface product (audience-facing fan cam + operator dashboard that queues winners for the jumbotron) tied to a specific team. QR → selfie → Claude Vision scoring across energy/Celtics-spirit/originality → AI-generated PA-announcer roast lines is a credible live-event flow. Real multimodal AI use, all three URLs present plus a slide deck. Held under the top three because the deploy is ngrok-hosted (transient rather than a permanent URL) and the use case is entertainment-only — no analytics or coaching value beyond the moment.
Instant Appeal
AI 7.8by kvarnik
Courtroom-style settler to help parents and fans resolve controversial sports calls. Two theatrical lawyers argue both sides; a warm-hearted judge ends the debate with a shareable verdict card.
sportsllmnext.jscourtroomAI judge: Most imaginative framing among the 8s — multi-agent LLM courtroom (two theatrical lawyers + a warm-hearted judge) settles controversial sports calls into a shareable verdict card. Sharp narrative pitch, all three URLs present (Loom + repo + Vercel deploy), and multi-agent persona-driven argument is a non-trivial AI pattern rather than a single prompt. Lands just below the other 8s because the use case is purely entertainment/social — there's no coaching, analytics, or live-event ops surface, so the ceiling on impact is lower even though the build is complete and well-shipped.
PulseLens Live: AI Subtitle-to-Insight Companion for Sports Streams
AI 7.0by Hanpeng Yang
PulseLens Live turns real-time stream subtitles into instant scene understanding, breaking-news style summaries, explainable term highlights, and timeline-grounded Q&A. It helps solo online viewers quickly understand what is happening and why it matters.
fan-experiencelive-subtitlessports-streamingai-qahackathonAI judge: PulseLens Live is a sharp, recognizably sports project: a real-time subtitle-to-insight companion for solo sports-stream viewers, with a concrete feature set (scene understanding, breaking-news summaries, explainable term highlights, and timeline-grounded Q&A). The strongest signal is its AI-native shape — subtitle summarization plus grounded Q&A is genuine LLM leverage, not a bolt-on — and it ships with both a public repo and a live deployed URL. The main weakness is the missing explainer video, which leaves the completeness picture short of the all-URLs bar; otherwise the pitch is well-scoped and ambitious for a 2-hour build.
Sub Verdict
AI 6.0An AI-powered tool that analyzes Premier League substitutions and delivers instant verdicts on whether a manager made the right call. Users select a match and substitution to get a Brilliant, Fine, Too Late, or Wrong Call verdict with tactical reasoning powered by Groq LLaMA.
premier-leaguefootballaitacticsstreamlitAI judge: Sharp, narrowly-scoped pitch: rate Premier League substitutions with a structured Brilliant/Fine/Too Late/Wrong Call verdict + tactical reasoning. Clear sports fit, sensible AI use (Groq LLaMA for tactical commentary), and a deployed Streamlit demo. Knocked down because videoUrl is empty — the brief required a short explainer, and the deployed app alone can't substitute for the demo. Solid concept that would land in the 7-8 range with a video.
Aesthetix Athlete Lab: AI Active Recovery & Bio-Tuner
AI 6.0by Ankit Yadav
A premium sports recovery dashboard utilizing the Vercel AI SDK and Groq. Features dynamic Generative UI, including an age-Karvonen EKG heart-rate zone monitor with a native Web Audio API lub-dub beat synthesizer, interactive front/back muscle heatmaps, and real-time parent-child state syncing.
next.jsvercel-ai-sdktailwind-cssweb-audio-apiekg-simulatorgenerative-uiAI judge: Technically ambitious pitch with strong AI leverage signal — Vercel AI SDK + Groq powering generative UI, plus a Karvonen-formula EKG zone monitor, muscle heatmaps, and Web Audio API lub-dub. Specific tech stack named, video present, sharp framing. Held back from the 7-8 tier because deployedUrl is empty — the brief weighted shipping live, and on a recovery dashboard the deployed demo is most of the proof. Solid submission that would have ranked clearly higher with a working URL.
Ski Buddy Matcher
AI 6.0This app is designed to help people find compatible ski buddy matches for ski trips! Not just compatible terrain but - Would these people actually enjoy spending twelve to forty-eight hours trapped in snow together? A ski weekend compresses roommate dynamics, travel dynamics, and activity dynamics into like forty eight hours
AI judge: Novel human-centered angle — match ski buddies on whether they'd actually enjoy 12–48 hours trapped together, not just terrain fit. All three required URLs are filled in (Loom + repo + Streamlit deploy). Loses points because the description trails off mid-thought and the AI leverage isn't called out in the meta — the matching engine is presumably LLM-backed but the pitch doesn't sell it. Solid niche submission that could land higher with a sharper, complete pitch.
Sideline IQ
AI 5.0by Prasanth
Sideline IQ is an AI-powered sports coaching assistant that helps coaches and analysts make smarter in-game decisions. It provides real-time insights, player performance analytics, and strategic recommendations — all from the sideline.
sportsaicoachinganalyticsnext.jsAI judge: Solid attempt with all three required URLs (Loom + repo + Vercel deploy) and a recognizable sports angle (in-game coaching assistant). Held back by a generic pitch: "AI-powered coaching assistant with real-time insights, performance analytics, and strategic recommendations" reads like a product brochure rather than a specific build — no named sport, no named data source, no concrete decision the system actually helps with, and no AI-leverage specifics beyond the "ai" tag. The PR also targets `develop` instead of the `sports-hack-2026-submissions` branch, which signals less attention to the contribution flow. Lands at the median of the pack.
BeatM ROAST MODE - AI Trash Talk Battle Cards for DFS Lineups
AI 4.0Enter two DFS lineups, get AI-generated trash talk battle cards. Claude roasts both sides with stat context, crowd predictions, and shareable cards. Built for BeatMs social-first DFS thesis.
beatmdfsai-roastsocialclaude-ainext-jsAI judge: Creative niche idea — AI-generated trash-talk battle cards for two DFS lineups, tied to a real-product thesis (BeatM) — and the pitch is specific enough to picture the UX. Loses heavily on completeness: both videoUrl AND deployedUrl are empty, so there's no way to verify the project actually shipped, and the brief explicitly required a live URL + explainer video. With either of those filled in this lands in the 6-7 range; with both missing it can't break into the solid tier.
FanDraft — AI-coached fantasy lineup builder
One user: casual DFS players. One problem: spending 45 minutes before tipoff pulling projections and scanning injury news. One product: voice-driven AI agent that builds tonight's NBA lineup under a $50K cap in 10 seconds, with per-pick reasoning grounded in live injury news. Built end-to-end in Cursor with OpenAI function calling over the NBA Stats API and ESPN injury feed.
nbafantasydfsagentsvoice-aiopenaiSportsHire AI — MCP-native sports job marketplace
by Arpan D
Structured seeker profiles and employer-weighted scoring criteria. MCP server + REST /api/tools/* with 10 agent tools (search, rank, apply). 367 Boston-area jobs (362 scraped from WorkInSports + curated demo). AI-judge docs: AGENTS.md, docs/AI_JUDGE.md. Vercel-ready Next.js deploy (docs/VERCEL.md). Freemium UI demo; agents get full API access with SPORTSHIRE_AI_JUDGE=1.
sportsmcpjobsai-agentsmatchingbostonThe Assist
by Calvin V. with Karlee Perpignant, Emily Castillo, Calvin V.
The Assist is a Celtics watch experience for first-time basketball viewers who feel lost during the game. A public Finals clip drives a synced overlay that explains what you are seeing on screen, what announcers mean, and rules in plain language, plus a live score strip, transcript rail, and confetti on big scores.
basketballcelticsfan-experiencereactviterookie-modeMatchRoom: Verified AI Coaching Room for High-Stakes Team Decisions
Every team drowns in data but starves for decisions it can trust. MatchRoom turns raw game data into a coach-ready brief through a staff of AI agents: a Scout proposes insights, a Skeptic attacks weak and small-sample claims, and only evidence-backed conclusions reach the coach, each linked to the proof. Baseball advance scouting is the wedge; the same evidence-gated engine scales to any high-stakes team that must decide fast and be right. No invented numbers.
baseballai-agentsscoutingverificationdecision-intelligencenext.jsScoutAI — International Soccer Scouting for USL & College Programs
by Daniel Chen
USL clubs and college programs can't afford Wyscout. ScoutAI uses a two-stage Claude pipeline with league-adjusted statistical projections to generate professional scouting reports on international players instantly.
soccerscoutingaiclaudesports-analyticsHeartbreak Weather: Boston Marathon Hill & Environmental Simulator
by I-Ju Lin
An interactive pacing and physiological analysis platform that models the 26.2-mile elevation profile of the Boston Marathon against real-time weather conditions, including extreme temperature waves, humidity, and wind speed vectors.
boston-marathonweather-modelpacing-strategygait-biomechanicsdata-analyticsRepRef — AI Wall Ball Judge for HYROX
by Jingren Ma & Shahza Saeed with Jingren Ma, Shahza Saeed
RepRef uses on-device pose tracking to judge HYROX wall balls objectively — verifying squat depth and ball-to-target height in real time — replacing inconsistent human rep judging. The iOS version uses ARKit body tracking for true metric 3D; a web prototype runs in any browser.
hyroxcomputer-visionarkitsports-techKinetiQ — Overhead Squat Form Analysis
by justinchan36
KinetiQ is a web app where athletes upload a video of an arm overhead squat. Computer vision and Gemini analyze movement to flag issues like knee valgus, ankle inflexibility, and weak core, then generate a correction plan and suggested workouts. Not deployed live; run locally via the README.
computer-visionform-analysisfastapinext.jsgeminiHoops AI — Basketball Injury Prevention & Performance Coach Console
Injury ends promising careers early and poor training optimization leaves potential unrealized. Hoops AI is a coach console that ingests player health signals, game stats, and live session metrics, then uses Claude with RAG-grounded context to flag injury risk and deliver actionable next-session recommendations — protecting careers and accelerating improvement.
basketballinjury-preventionclauderagfastapireactInjuryIQ — Athlete Recovery Risk Predictor
Athletic trainers face a high-stakes daily call: should this athlete practice today? InjuryIQ takes a 5-field morning check-in (training load, sleep, soreness, days since rest, prior injuries) and returns a calibrated injury-risk score with the top 3 driving factors for that athlete today. Built on a star-schema data model with a logistic regression trained on 1,500 athlete-days.
sports-techathlete-performancemlstreamlitinjury-preventionRockie
by n-star-dot with Philip, Nghia, Duc
AI-powered rock climbing performance app that uses computer vision to analyze climbing technique, track progression, and deliver personalized coaching insights.
computer-visionreact-nativeexpoATHLIX AI
AI-powered financial intelligence terminal for professional athletes. Predicts injury-linked earning decline, career instability, and retirement collapse risk using interactive simulations and a streaming AI analyst.
sportsfinanceainextjsrechartsterminalFanPulse FIFA AI
FanPulse FIFA AI helps football fans rediscover past matches through personalized AI storytelling, top highlights, hidden insights, trivia, and shareable social content. The MVP uses a 3-agent pipeline and supports OpenAI with mock fallback; it is not deployed yet.
aifootballfan-engagementsportsfastapiopenaiCoachCam AI - A free-throw form analyzer
by Shivu & Prathamesh with Shivu, Prathmesh
CoachCam AI helps beginner basketball players improve their free throw form by analyzing a short phone video and giving instant visual + coaching feedback.
basketballpose-estimationnext.jsCricket IPL Career Progression Visualizer
Static export of an IPL player career progression analysis notebook (2008–2025) with downloadable CSV datasets for batting, bowling, and player master records.
cricketiplstatsvisualizationnotebooksquadpulse
by squadpulse with lvcasmadeit, sowndyjay
affordable health tracking metrics with consumer hardware
ScoutAI — AI Sports Intelligence Agent with Live Web Search
ScoutAI is a multi-mode sports intelligence platform that gives coaches and fans access to tactical analysis powered by real-time web search. It features opponent scouting, a conversational live intel chat, head-to-head player matchup comparisons, and a tactical game plan builder, all using live data.
ai-agentsports-analyticsweb-searchopenai-gpt4oreal-time-datacoachingFootball IQ — train your game sense on real moments
by Yaman Bicer with Ahmet Sukru Kilic
Football IQ is a mobile decision-training game built on real Champions League moments. Each clip auto-pauses at the decision point — pick what the player should do, then see what actually happened alongside a Veo-generated "alternate timeline" of how your wrong choice would have played out. Career mode filters scenarios by position, and you can upload your own clip to be analyzed and trained on.
soccerdecision-trainingveoreactviteCoachLens Court
by zukhriddingit with atharvg-BU
CoachLens Court is an AI micro-coach for tennis forehands. It crops the core swing, links pose evidence to one coaching fix, drill, and question, and adds an AI review/share flow.
tennisai-coachsports-techmediapipereactgemini