Handbook
🚀 About the Hackathon
In-Person & Virtual: August 30–31, 2025 — 24‑hour global AI sprint.
Join us in Paris or fully online.
👥 Who Should Join
- Builders who ship running code (no pitch-only projects)
- Individuals or teams of up to 4
- Technical roles encouraged: ML engineers, full-stack, data scientists, product/UX
📜 Rules & Deliverables
- Teams of max 4; work must be created or substantially extended during the hackathon
-
Submit a
project including:
- Project Name & Description
- Products & Tools Used
- Team Contributions
- Optionally, link a 2-minute demo video (deployed service or local demo)
- Submit a GitHub repo (public or private with access), README, and optional project website
Judging Criteria: 🔎 Detailed Scoring
| Criterion | Weight |
|---|---|
| Technical | 35% |
| Innovation | 25% |
| Presentation | 20% |
| Assets | 20% |
| Bonus (survey & feedback) | +10% |
Submission Checklist (before Aug 31, 3:00 PM)
- GitHub repo + README with run instructions
- Deployed demo link or local test instructions
- 1‑page slides or README problem/impact summary
- 2‑minute demo video (recommended)
- Required credentials/access for judges
- Short note on Responsible AI & limitations
🤝 Sponsors & Organizers
- Mentorship from MIT, Harvard, and industry leads — ask questions on the message board
- Free compute credits & platform access —
- Key sponsors & partners: SarApi, Schoolab, FinTech Dev Night, Lovable,
Aethos.org, SarAI,
Conduit.ai
🧑🏫 Mentors & Office Hours
Sign up for mentor sessions and Bring specific asks to maximize feedback (e.g., “debug agent loop”, “scale retrieval”)
Highlighted Mentors:
Lodovico BENVENUTI (Early Stage…)
Alaa ABDULRIDHA (Search…)
Vrinda BHATIA (MCP…)
Ammar EKBOTE (PaaS…)
Sujata SRIDHARAN (Fintech…)
Abdoulaye DOUCOURE (Seed for Good…)
🛠 Suggested Tech Stack & Tooling
- LLM / Models: OpenAI, Anthropic, Mistral, Cohere, Hugging Face, Google Gemini
- Integration Frameworks: LangChain, LlamaIndex, Haystack
- Agent Frameworks: LangChain Agents, AutoGPT/BabyAGI (be cautious of safety/loops)
- Vector DBs / Retrieval: Weaviate, Pinecone, Milvus
- MLOps & Monitoring: Weights & Biases, MLflow, logging + health endpoints
- Hosting / Deployment: Vercel, Koyeb, Heroku, Cloud Run, Docker + ngrok
- In-browser / Edge LLMs: WebLLM, Ollama, on-device inference
📁 Starter Repo Structure (Suggested)
/README.md → run instructions, API keys
/app → frontend/demo
/backend → model integration, APIs
/data → sample data & indexing
/infra → Dockerfiles & deploy scripts
/presentation → slides + demo video links
💡 Project Idea Prompts
- Ethical AI: transparency, fairness, privacy-preserving solutions
- Agentic AI: autonomous planning and execution
- In-browser personal LLM with local retrieval & privacy
- Multimodal search: image+text retrieval for e-commerce or research
- FinTech: explainable models for transaction categorization, compliance, or support agents
- Conversational RAG systems with evaluation pipeline
⚡ Hacking Tips & Priorities
- Scope small: one happy path + working demo > many features
- Start with tiny or synthetic datasets to prove flow
- Use managed services to avoid infrastructure time sinks
- Containerize early for consistent demos
- Record short guided demo video in case of network issues
🛡 Support & Logistics
- Venue / registration questions:
- Remote participants: join Discord to connect with teammates
🔐 Responsible AI & Data Privacy
- Explicitly state data provenance and model usage in README & slides
- Avoid uploading PII without consent
- Include limitations & safety note in submission
📬 Contacts & Help
- Event page & RSVP:
- Organizers (Paris):
Abdoulaye DOUCOURE +
- Urgent ops issues:
💡 Build fast, demo clearly, and have fun!
We’re watching for creative, reproducible projects that ship working code.