AgentsIRL: Multimodal AI Hackathon with Google Cloud Run, Mistral, and NVIDIA GPUs [AI Tinkerers - Paris]
Cities (231) ▾

AgentsIRL: Multimodal AI Hackathon with Google Cloud Run, Mistral, and NVIDIA GPUs

Oct
11
Saturday
Saturday, October 11th, 2025 • 9AM to 9PM (CEST)
Address Info
Available on RSVP acceptance

Event Ended

This event has already taken place.

Gallery Hackathon Portal
Attendees 203+ registered
Attendees include engineers from Google, Meta, and Mistral AI, with deep expertise in Python, agentic RAG, MLOps, and high-performance computing.

🧠 AgentsIRL: Multimodal AI Hackathon

(Banner) This is a digital banner promoting the "Agents IRL Multimodal AI Hackathon" scheduled for October 11 in Paris. Text: AI TINKERERS October 11 | Paris Agents IRL Multimodal AI Hackathon ft. Google Cloud Run, NVIDIA, & Mistral NVIDIA Google Cloud Mistral AI modern, bold, high contrast, slightly retro/glitch aesthetic | Colors: #000000, #ff4a00, #0033cc, #ffffff Note: This image functions as a promotional poster or advertisement, using large text and structured layout to announce an event, which categorizes it as a banner.

**ft. Google Cloud Run, NVIDIA, and Mistral **

Build agents that see, hear, and think — then deploy them.

✨ Event Overview

Join us for AgentsIRL, a one-day technical sprint where AI builders will prototype real-world, multimodal agents that perceive and respond to audio-visual input — all deployed via Google Cloud Run, and accelerated with NVIDIA GPUs.

(Logo) The image features the logos of NVIDIA and Google Cloud placed side-by-side on a solid black background, typically used to signify a corporate partnership or collaboration. Text: NVIDIA. Google Cloud Colors: #000000, #FFFFFF, #76B900, #4285F4 Note: The image is a graphic display consisting solely of two distinct, recognizable corporate logos (NVIDIA and Google Cloud) placed together for co-branding purposes.

Developers love Cloud Run, Google Cloud’s serverless runtime, for its simplicity, flexibility, and scalability. Now, with NVIDIA L4 GPU support, Cloud Run offers a powerful environment for AI inference, perfectly suited for running a wide array of lightweight open models. This includes popular Open LLMs such as Gemma 3 1B/4B/12B, DeepSeek-r1 7B, and Llama 2 7B, as well as powerful open foundation models like Stable Diffusion XL (SDXL), Whisper, YOLOv8, and SAM.

(Photo) An indoor event shows a speaker presenting to a large audience, with many attendees visible from behind. A large screen displays text and an image of a futuristic London skyline, suggesting a tech-related event. Text: AI Tinkerers London Ultimate Agents Hackathon Date: [unreadable] / Time: [unreadable] / Location: [unreadable] / Register: [unreadable] A speaker giving a presentation to an audience at an event. | Indoor, likely an event hall or conference room. | A large group of people, a man speaking at the front, a large screen displaying event information and a futuristic city image, potted plants, and light green curtains. | Candid, event photography, wide shot. | Colors: #EAEAEA, #829C74, #2F2F2F, #D0D0D0, #000000 Note: The image is a direct capture of a real-world scene, featuring people, an indoor environment, and a presentation, which are characteristics of a photograph.
Photo: Our Hackathon in London last week

☝ Register now. Space is limited. ☝

🧪 Hackathon Ideas:

Teams can build agents that have great real life use cases. Examples:

Inspired by recent research directions in spatial reasoning and AV-LLMs (like SAVVY), this hackathon pushes teams to create agents that interpret real-world scenes and answer grounded, temporal questions such as:

“Where is the speaker located?”
“What caused the glass to fall?”
“Who just entered the room, and what did they say?”

Whether you’re working with Whisper, YOLOv8, SAM, challenge yourself to fuse audio, video, and inference into live demos that can run anywhere, anytime, and at scale
  • Interpret live AV feeds from webcams or mobile devices
  • Analyze videos for events, causes, and speaker locations Simulate embodied intelligence for games, security, or accessibility
  • Multi-Image Comparison and Analysis
  • Interactive E-commerce and Retail Experiences

Bonus challenges for:
  • Most elegant Cloud Run deployment
  • Best use of GPU acceleration
  • Most useful real-world agent (robotics, accessibility, spatial UI, etc.)

☝ Register now. Space is limited. ☝

🗓 When & Where

🗓Saturday Oct 11 📍 In-person, Paris (venue on RSVP)
🧑‍💻 Limited slots, technical applications only
Time Activity  
  8:00 AM Doors open · Registration
  9:00 AM Opening and Cloud Run Serverless GPU walkthrough
  10:00 AM Team-formation lock-in Session
  12:00 PM Lunch
  1:00 PM – 6:00 PM Mentoring available (sponsors & judges)
  6:00 PM Dinner break
  7:15 PM Project submission deadline · Judging starts ( live demos from team who make the final rounds
  8:15 PM Awards & closing ceremony
  9:00 PM Doors close

🔗 RSVP

Get early access to the full event brief, starter containers, and judge lineup.
This event will fill quickly — RSVP now.

📊 AI Tinkerers Paris Stats

  • Attendees: This exclusive community of 4,082 subscribers comprises elite technical professionals, with 42% specializing in AI/ML research, 31% in full-stack software engineering, and 27% in technical leadership. Members possess deep expertise in PyTorch, LLMs, and agentic architectures. Notably, the group features prominent open-source maintainers from Hugging Face and graduates from top-tier institutions like École Polytechnique.
  • Companies Represented: Featuring tech giants like Google, Meta, and Microsoft, alongside open-source and AI leaders such as Hugging Face, Mistral AI, LangWatch, and Black Forest Labs, and innovative startups like ElevenLabs, Docker, Datadog, Algolia, and Zoom, and more
  • Demos: Across 130 submissions, 83 demos have been presented at AI Tinkerers - Paris. The most exciting themes have focused on agentic production workflows, knowledge-graph-driven reasoning (GraphRAG/context graphs), and system architectures for scalable, reliable deployments. Technically, structured outputs and tool/function calling, cost/latency reductions, eval/benchmarking, and multimodal/local execution have been repeatedly explored.
  • Testimonials:
    “Overall, it was high-quality content and it made me want to come back for the next ones!”
    — Product manager, AI community

☝ Register now. Space is limited. ☝

What’s Next

If you’re planning to sponsor future events or expand participation, you can find sponsor opportunities at: https://paris.aitinkerers.org/sponsors

(Photo) The image captures a view from behind an audience attending a presentation, with a projector screen visible in the background displaying text and a logo. Many individuals are seated, looking towards the screen, suggesting a conference or lecture setting. Text: https://www.linkedin.com/in/etiennebcp/ went to the store and bought 1 type of milk. Their purchases cost $1 and $2 respectively. Then Alice bought a bike. AI Tinkerers, Sep 3rd 2024 AI TINKERERS NuMind BRING US YOUR MOONSHOTS audience, presentation, people, event | indoor, likely a conference room or auditorium | multiple people viewed from behind, projector screen with text and a logo, text on a person's shirt | candid, documentary | Colors: #2C354E, #6B8BA6, #A07452, #F0F0F0, #5C4336 Note: The image is a realistic capture of a moment or scene, depicting people and an environment as they appear in real life, which is characteristic of a photograph.
Photo: AI Tinkerers - Paris

Ready for more?

Check out other posts from this blog.

View all posts

Contact Organizers

Questions? We're here to help.

Events Gallery FAQ Sponsor