Hacking with GenAI 101

Who am I

Bauke Brenninkmeijer

✓ CS @Radboud University
✓ 5 Years Data Science @ABN AMRO
✓ Transitioned to GenAI last year @ING
✓ MLOps Community Chapter Lead Amsterdam

Start with why

GenAI will change everything

Where are we going…

the chatgpt launch 26 months ago was one of the craziest viral moments i'd ever seen, and we added one million users in five days.

we added one million users in the last hour.
— Sam Altman (@sama) March 31, 2025

The Ghibli AI Controversy

“An Insult to Life Itself”

— Hayao Miyazaki

The Landscape

Has Ghiblified a photo?

Use AI daily?

Used an AI coding assistant

Used deep research

AI Adoption Rate: Firm Sizes

Law Adoption Rate: Individual vs. Firm

Why GenAI is different

1. Generate vs. Evaluate

2. Multi-modality

3. Generality

The Technology

To be precise: predict probabilities of the next word

Tokens != words

Temperature

Context windows

How much stuff you can put it.

Context windows

🎮 Choose Your LLM!

🔄 I/O Capabilities

📥 Input modalities: text, image, audio, voice
📝 Output modalities: text, image, voice
🌍 Multilingual capabilities

⚙️ Technical Features

🧠 Reasoning
🛠 Tool Calling
🔢 Structured Output

📊 Practical Factors

📏 Maximum context window size
💰 Price per token (input/output)
⚡ Response generation speed

Reasoning Models

💰 What is the cheapest model to use?

FREE

🌟 Gemini

✓ Gemini 2.5 Flash: 10 calls/min, 250k tokens/day
✓ Gemini 2.0 Flash: 15 calls/min, 1M tokens/day
✓ Gemini 2.0 Flash Lite: 30 calls/min, 1M tokens/day

FREE

⚡ Nvidia (build.nvidia.com)

✓ OSS models: (Llama, Mistral, DeepSeek)
✓ Limits: 40 calls/min, unlimited tokens
⚠ Limited framework integration
⚠ Forced usage of Nvidia Langhchain

€10

GPT-4.1-mini Budget Example

10k input tokens + 1k output tokens ≈ 2000 API calls

Should be sufficient for a hackathon

The Tools

Visual AI Frameworks

✦ LangFlow
✦ FlowiseAI
✦ n8n

Tools

🔍 Web Search

API Searches
News Analysis

🕸️ Web Scraping

Content Extraction
Browser Automation

📚 RAG Systems

Document Retrieval
Context Management

🗄️ Vector DBs

Similarity Search
Embedding Storage

🔌 API Clients

REST/GraphQL
Authentication

🗃️ Database

SQL/NoSQL
Data Querying

💻 Code Gen

Code Analysis
Autocompletion

🛠️ Dev Tooling

Git Operations
Execution Envs

⚙️ Shell Access

Command Execution
System Integration

📂 File System

File Operations
Data Processing

📧 Messaging

Email/SMS
Chat Platforms

🔔 Notifications

Push/Webhooks
Social Media

🎨 Image Tools

Generation
Analysis/OCR

🔊 Audio/Video

Speech Processing
Media Analysis

🧠 Reasoning

Chain-of-Thought
Logical Analysis

🧩 Planning

Goal Decomposition
Self-Reflection

Tools

Programmic AI Frameworks

🔗 LangChain 🚩

🦙 LlamaIndex

🌾 Haystack

PydanticAI

Self-Hosting LLMs

Ollama

vllm

LM Studio

Illustration of self-hosted LLM infrastructure

Could you self-host?

💻

MacBook

Perhaps

🖥️

Other Laptops

Not really

☁️

Cloud

Yes, but not for the hackathon

Should you self-host?

🏆

For the hackathon

🔒

Security & Privacy Critical

Yes

Development Tools

📝 VS Code

👨‍💻 Github Copilot

📊 Cline

🦘 Roo Code

💻 IDE

Cursor

🌊 Windsurf

⌨️ CLI

🔧 Aider

🤖 Claude coder ⚠️Only Claude Sonnet

🌐 Web

🔥 Firebase Studio

❤️ Lovable

Biggest Challenge

Specificity

⛰️

GenAI will fail when given tasks that are:

✘ Too complex without breakdown
✘ Ambiguous in requirements
✓ Break down complex tasks
✓ Be specific in instructions

The Patterns

Effective Prompting

• Few-shot prompting
• Chain of Thought
• Self-Consistency

Use LLMs to prompt LLMs

Development Approach

Steps

🧩

Task Decomposition

🚀

Implementation

🧪

Lots of Tests

Cost Management Strategies

⚡

Caching

📦

Batching

📊

Token Usage

MCP: The Protocol That Connects Worlds

"MCP is an open protocol that enables seamless integration between LLM applications and external data sources and tools."

— Anthropic

Recommendation

First check if an MCP server exists already.

The Agents

Types of Agents: Single Agent System

Multi Agent Systems

Hierarchical

Group

AI Agents Frameworks

🔄

LangGraph

👥

CrewAI

🤖

Autogen

🧠

Agents SDK

🛠️

Agent Development Kit

Exponential growth

Domains to focus on

🚀 Hackathon Success Blueprint

Choose the Right Model

Validate that free tier models are sufficient for your task.

Use Existing Tools

Leverage frameworks (LangChain) and productivity tools (Cursor, GitHub Copilot)

vibe code 80%, refine 20%

Check for MCP Servers

Integrate external tools via existing MCP servers before building your own

Think Human-Centered

Focus on solving a real problem with great UI/UX design