ExecuTorch Hackathon — Build On-Device AI on Snapdragon

📍 The Web Data Loft by Bright Data · 625 2nd Street, San Francisco, CA 94107 📅 June 27–28, 2026 · On-site only · Apply by June 15

Two days. Real Snapdragon hardware. AI that runs entirely on the phone in your hand.

The ExecuTorch Hackathon brings 150 builders to San Francisco to take PyTorch models on-device with ExecuTorch and ship something real on Samsung Galaxy S25 Ultra devices powered by Snapdragon — no cloud round-trips, no latency, no compromise on privacy.

This is a hands-on, in-person build hosted by Qualcomm, Meta, GitHub, and the PyTorch Foundation, with mentors from each on the floor for the full two days. You bring an idea and a team of 3–5; we bring the devices, the experts, and the tooling.

What you'll build

Working in teams of 3–5, you'll design, build, and demo a real, user-facing application that runs locally on a Snapdragon-powered mobile device using PyTorch and ExecuTorch. The strongest projects make a clear case for why the AI needs to run on-device — lower latency, offline capability, privacy-sensitive processing, energy efficiency, or a smoother real-time mobile experience.

Suggested directions:

  • Computer vision — real-time detection, recognition, scene understanding, accessibility, or camera-based tools

  • Multimodal assistants — mobile assistants that combine text, images, audio, and context

  • Generative AI on mobile — local generation, summarization, transformation, creative workflows

  • Privacy-first AI — apps where keeping data on the device is the whole point

  • Real-time edge intelligence — fast, low-latency tools for everyday mobile use

Who should attend

This event is built for people who want hands-on time with on-device AI, not a lecture about it. You'll fit right in if you're a:

  • PyTorch / ML engineer who wants to take models from PyTorch to ExecuTorch and run them on real hardware

  • Mobile or Android engineer curious about deploying AI locally on Snapdragon

  • Optimization-minded developer working on quantization, NPU utilization, latency, and efficiency

  • Computer vision, multimodal, or generative AI builder who wants a mobile target

  • Founder or product builder prototyping privacy-first or offline-capable AI products

  • Bay Area developer who wants two days of focused building alongside Qualcomm, Meta, and GitHub engineers

Apply with an idea. If you need teammates, we will help you find them.

Agenda (all times PDT / San Francisco local)

Saturday, June 27

  • 10:30 AM — Check-in and device loaning

  • 12:00 PM — Kick-off + master class: PyTorch, ExecuTorch, and recommended build tools

  • 1:00 PM — Hacking begins

  • 5:00 PM — Dinner

  • 8:00 PM — Doors close for the night

Sunday, June 28

  • 9:00 AM — Doors reopen, hacking continues

  • 11:30 AM — Lunch

  • 1:00 PM — Submission deadline

  • 1:00–4:00 PM — Team demos (5 minutes per team)

  • 4:00–4:15 PM — Device collection

  • 4:00–4:45 PM — Judging & evaluation

  • 4:45–5:00 PM — Winners announced

  • 5:00–7:00 PM — Social hour & celebration 🎉

Prizes

🥇 Top Award (selected by judges)

  • Meta Quest 3 512GB Mixed-Reality Headset for every team member

  • Qualcomm DevRel support to finish your app and publish it on the App Store

  • Blog + live stream feature to showcase your project

🗳️ Team's Choice Award (selected by participant vote)

  • Ray-Ban Meta AI Glasses for every team member

  • Qualcomm DevRel support to finish your app and publish it on the App Store

  • Blog + live stream feature to showcase your project

💻 Copilot-Powered Build Award (selected by GitHub judges)

  • Up to 6 months of GitHub Copilot Pro+ and GitHub for Startups credits for teams who use Copilot creatively across the build

Note: Top Award winners are not eligible for the Team's Choice Award.

Hosts & partners

Qualcomm — Maker of the Snapdragon platform powering the devices. The Qualcomm AI Hub, sample apps, and on-site DevRel engineers are your fast track to running models on the NPU.

Meta — Creators of PyTorch and ExecuTorch, the on-device runtime at the heart of this hackathon, purpose-built to take PyTorch models to efficient mobile execution.

GitHub — Bring Copilot as your AI pair programmer and compete for the Copilot-Powered Build Award, judged by the GitHub team.

PyTorch Foundation — Steward of the open-source PyTorch ecosystem the entire event is built on.

Samsung (hardware partner) — Shortlisted teams build and test on Snapdragon-powered Samsung Galaxy S25 Ultra devices.

Bright Data — gives builders access to the infrastructure needed to collect, reason over, and act on live public web data at scale.

Speakers, mentors & judges

  • Lauren Lunde — Senior Program Manager, Qualcomm.

  • Andrew Caples — Business Development Manager, Meta.

Industry mentors on the floor across both days to keep your team unblocked.

What you'll work with

ExecuTorch · PyTorch · Snapdragon NPU · Samsung Galaxy S25 Ultra · Qualcomm AI Hub · GitHub Copilot, plus a full set of sample apps and developer resources shared ahead of the event.

The venue — The Web Data Loft by Bright Data

We're hacking at The Web Data Loft — a dedicated builder and event space in San Francisco's South Park / SoMa neighborhood, and already a gathering spot for the city's AI and agent-building community. It's the new SF home of Bright Data, the web data infrastructure powering AI and BI for 20,000+ companies.

📍 625 2nd Street, San Francisco, CA 94107 — walkable from South Beach and the surrounding neighborhoods.

Presented by lablab.ai

lablab.ai is the global platform behind 100+ AI hackathons and a community of 250,000+ builders shipping with frontier AI. We turn weekends into working prototypes — and prototypes into companies.

Apply by June 15, 2026. On-site in San Francisco only. Build where it matters most — on-device.

  • luma
  • trusted-calendar