AI Project Concept Template for Windesheim/VCH

All projects follow Agile principles with iterative development, regular sprint cycles, and continuous stakeholder feedback.

PROJECT CONCEPT

Basic Information

Project Title: Chatbot for VCH Website

Project Slug: chatbot-vch-website

Concept Status: brainstorm

Category: student-project / internal-tool

GitHub Repository: [TO FILL - Does this exist yet?]

THE 3 MINUTE RULE PITCH

1. What is it?

An AI chatbot on the VCH website that answers questions about projects, research, courses, and supply chain topics.

2. How does it work?

[TO FILL - RAG (Retrieval-Augmented Generation) using VCH website content + project docs? Fine-tuned model? Which LLM provider?]

3. Are you sure?

Chatbots for educational websites are proven. The challenge is making it knowledgeable about VCH-specific content and keeping it updated.

4. Can you do it?

[TO FILL - Team has LLM experience? Infrastructure for hosting?]

5. What’s the value?**

Visitors can ask questions 24/7 about VCH projects, research areas, how to collaborate, etc. Reduces email load. Helps potential students/partners learn about VCH.

6. Are there any risks?

• Chatbot gives incorrect information about VCH
• API costs for LLM usage
• Needs constant updating as VCH content changes
• May answer questions about sensitive topics incorrectly
• Maintenance burden

The Problem

What problem does this solve? People have questions about VCH projects, collaboration opportunities, research areas, but have to dig through the website or send emails. Response time is slow.

Who has this problem?

• Potential student collaborators
• Partner companies exploring collaboration
• Researchers looking for VCH expertise
• General visitors wanting to understand VCH

The AI Solution

What AI/ML technique would you use?

• RAG (Retrieval-Augmented Generation): Embed VCH content, retrieve relevant docs, use LLM to answer
• Possibly fine-tuning on VCH-specific Q&A pairs
• LLM: OpenAI GPT, Anthropic Claude, or open-source (Llama, Mistral)

What data would you need?

• All VCH website content
• Project descriptions and documentation
• Research publications
• FAQ content
• Course descriptions
• Common questions from emails/Discord
• Data availability: We have this, just need to structure it

Expected output/deliverable: Chat widget embedded on VCH website, responds to questions about VCH projects and supply chain research, links to relevant pages.

FEASIBILITY CHECK

Technical Difficulty: Medium

Required Skills:

• LLM/RAG implementation (LangChain, LlamaIndex)
• Frontend integration (JavaScript chat widget)
• Backend API (Flask/FastAPI)
• Vector database (Pinecone, Weaviate, or pgvector)
• Prompt engineering

Resources Needed:

• Computing: Cloud hosting for API, vector database
• Data: VCH website scraping/export, project docs
• People: 1-2 students with AI experience
• Budget: LLM API costs (€50-200/month depending on usage), hosting (€20-50/month)

TIMELINE & MILESTONES

Total Estimated Time: 3-4 months

Suggested Phases:

1. Content collection and embedding (2 weeks)
2. RAG pipeline development (3 weeks)
3. Chat widget frontend (2 weeks)
4. Testing and refinement (2 weeks)
5. Deployment and monitoring (1 week)

[TO FILL - Detailed sprint breakdown]

STRATEGIC FIT

Student Learning Value: High - Students learn RAG, LLM deployment, prompt engineering, practical AI application.

Reusability: High - Could be template for other educational institution chatbots.

VCH Goal Alignment: Improves accessibility of VCH research, helps onboard new collaborators, demonstrates AI capability.

SUCCESS CRITERIA

Minimum Viable Product (MVP): Chatbot can answer 10 common questions about VCH accurately (who are you, what projects, how to collaborate, etc.)

Success Metrics:

• Accuracy: >85% of answers are correct and helpful
• Usage: 50+ questions per month
• Reduces email inquiries by 20%
• User feedback: >70% find it helpful

Stretch Goals:

• Multilingual support (Dutch + English)
• Integration with Discord bot
• Can answer technical supply chain questions, not just VCH info
• Conversational memory (remembers context within session)

NOTES FOR COMPLETION

Key Decisions Needed:

• Which LLM provider? (OpenAI cheapest, Claude best quality, open-source no API costs)
• Where to host? (Vercel, AWS, university servers?)
• What content sources to include beyond website?
• How to handle questions it can’t answer?
• Privacy considerations for chat logs?

Similar Projects to Reference:

• Educational chatbots (university websites)
• Documentation chatbots (Cursor, GitHub Copilot Docs)
• RAG implementations (LangChain tutorials)

[TO FILL - Full timeline, team assignments, tech stack decisions, budget approval]