​Join us for an intensive and inspiring weekend at the Holistic AI Hackathon 2025, where 40 teams will explore the next frontier of Agentic AI - intelligent systems that reason, plan, and act autonomously.

More: https://hackathon.holisticai.com/

​💡 What’s it about?

​Participants will select one of three technical tracks, each focused on a distinct challenge area within Agentic AI. The hackathon emphasizes creativity, collaboration, and real-world impact — blending cutting-edge research with practical innovation.

⚡ Tracks:

​Track A - Agent Ironman - “Robust, efficient, real-world agents”

  • ​Build agents that don’t break.

​Track B - Agent Glass Box - “Transparent, explainable agent behavior”

  • ​Follow the trajectory. Understand the behavior.

​Track C - Dear Grandma - “Red-team to safeguard AI”

  • ​Attack like a red-team. Defend like a guardian.


Hackathon Sponsors

Prizes

$65,000 in prizes
Prize Pots
$40,000 in cash
1 winner

Prize Pots
$25,000 in cash
1 winner

AWS + Valyu Credits and other.

Devpost Achievements

Submitting to this hackathon could earn you:

Judges

Emre Kazim

Emre Kazim
Co-Founder & Co-CEO @ Holistic AI

Graca Carvalho

Graca Carvalho
Director - UCL Centre for Digital Innovation @ University College London

Giuseppe Battista

Giuseppe Battista
Senior Solutions Architect @ AWS

John Adelana

John Adelana
Senior Solutions Architect @ AWS

César Ortega Quintero

César Ortega Quintero
Expert Data Scientist @ MAPFRE

Sergio Correa

Sergio Correa
Founder @ BMW Q✦Lab

Tigmanshu Bhatnagar

Tigmanshu Bhatnagar
Assistant Professor @ University College London

Hirsh Pithadia

Hirsh Pithadia
Co-founder, CEO @ Valyu

Alexander Ng

Alexander Ng
Co-Founder, Head of AI @ Valyu

Harvey Yorke

Harvey Yorke
Co-Founder, CTO @ Valyu

Daan Ferdinandusse

Daan Ferdinandusse
Founder @ Entrepreneurs First

Raj Patel

Raj Patel
AI Transformation Lead @ Holistic AI

Zekun Wu

Zekun Wu
Senior Research Scientist, Lead in Agentic AI @ Holistic AI

Kleyton Costa

Kleyton Costa
Research Scientist @ Holistic AI

Seonglae Cho

Seonglae Cho
Research Engineer @ Holistic AI

Rahul Patel

Rahul Patel
Performance Marketing Lead & LLM Security Testing, Holistic AI

Russell Bennett

Russell Bennett
Solutions Architect @ AWS

Judging Criteria

  • Technical Excellence (Overall Track)
    Code quality, innovation, and craftsmanship. Higher stars for elegant architecture, breakthrough solutions, and production-grade implementation. Show technical depth and creative engineering.
  • Poster & Presentation (Overall Track)
    Visual storytelling and reproducibility. Submit A3-A4 PDF poster + GitHub repo. Higher stars for clear communication, comprehensive docs, and effortless reproduction. Make your work shine.
  • Real-world Impact (Overall Track)
    Practical value and adoption potential. Higher stars for solving real challenges, demonstrating clear use cases, and showing transformative insights. Build something that matters.
  • Performance Optimization (Track A: Agent Iron Man)
    Engineer for speed and efficiency. Optimize latency, token usage, cost, and carbon footprint. Higher stars for measurable gains and production-ready monitoring. Build agents that scale.
  • Robustness & Reliability (Track A: Agent Iron Man)
    Harden against chaos. Handle prompt attacks, broken tools, and edge cases gracefully. Higher stars for resilient error recovery and production endurance. Build agents that don't break.
  • Observability Implementation (Track B: Agent Glass Box)
    Follow the trajectory. Capture every decision, memory update, and tool interaction. Higher stars for comprehensive traces with LangSmith and real-time visibility. Make autonomy transparent.
  • Explainability & Transparency (Track B: Agent Glass Box)
    Understand the behavior. Reveal reasoning chains and decision paths. Higher stars for human-interpretable explanations and true auditability. Turn opacity into clarity.
  • Attack Discovery & Severity (Track C: Dear Grandma)
    Uncover the cracks. Discover novel attacks with high severity: jailbreaks, prompt injection, tool misuse, data exfiltration. Higher stars for breakthrough vulnerabilities. Attack like a red-team.
  • Methodology & Impact Scope (Track C: Dear Grandma)
    How rigorous is your testing? Automate evaluation with clear metrics (LLM-as-Judge, rule-based, or multi-method validation). Find systemic vulnerabilities. Higher stars for comprehensive methodology and broad impact.

Questions? Email the hackathon manager

Tell your friends

Hackathon sponsors

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.