Eli the EtherFi Tutor

Inspiration

When exploring EtherFi—a protocol managing $8 billion in TVL—I realized the biggest barrier wasn't technology, it was education. Terms like "liquid staking," "restaking," and "EigenLayer" were intimidating even for crypto-curious users.

The problem: Most DeFi docs are written for experts, beginners have nowhere to ask "dumb questions," and static documentation can't adapt to different learning styles.

The solution: Claude AI as the perfect tutor—patient, knowledgeable, and able to explain concepts at any level. Combined with real-time data and interactive tools, Eli makes DeFi education accessible to everyone.


What it does

Eli is an AI-powered educational platform that makes EtherFi and DeFi concepts accessible through:

Intelligent Chat Interface

  • Natural language questions answered by Claude Sonnet 4.5
  • Adaptive responses based on expertise level

** Dual Learning Modes**

  • Beginner: Simple explanations with real-world analogies
  • Developer: Technical details and smart contract concepts

** Live Market Data**

  • Real-time ETH pricing (CoinGecko)
  • Current EtherFi TVL (DefiLlama)

** Interactive Staking Calculator**

  • Calculate rewards for any ETH amount
  • Automatic USD conversion
  • Formula: $\text{Yearly Reward} = \text{ETH Amount} \times \frac{\text{APR}}{100}$

** Knowledge Quiz System**

  • Multiple-choice questions
  • Instant feedback with explanations
  • Score tracking and achievements

How we built it

Tech Stack:

  • Frontend: Streamlit for rapid Python-based UI
  • AI: Claude Sonnet 4.5 via Anthropic API
  • Data: CoinGecko API (ETH price) + DefiLlama API (TVL)
  • Security: python-dotenv for environment variables

Development Process:

  1. MVP: Basic chat interface with Claude
  2. Context: Added EtherFi domain knowledge to system prompts
  3. Data: Integrated live pricing and protocol stats
  4. Features: Built calculator and quiz system
  5. Deploy: Secured keys, error handling, deployed to Streamlit Cloud

Key Implementation:

# Adaptive AI responses
prompt = f"""You are Eli, a friendly AI tutor specializing in EtherFi.
Mode: {explain_mode}
- Beginner: Use analogies, avoid jargon
- Developer: Include technical details
Question: {user_question}"""

# Live data with error handling
def get_eth_price():
    try:
        response = requests.get("https://api.coingecko.com/...")
        return response.json()["ethereum"]["usd"]
    except:
        return None

Challenges we ran into

1. API Type Inconsistency

  • DefiLlama sometimes returned TVL as list vs. number
  • Fixed with type checking: if isinstance(tvl, list): tvl = tvl[-1]

2. Prompt Engineering

  • Initial responses were too technical or vague
  • Solution: Detailed system prompts with context and teaching style

3. State Management

  • Quiz scores reset on page interactions
  • Used st.session_state for persistence

4. Security

  • Initially hardcoded API key (bad!)
  • Implemented proper environment variables with .gitignore

Accomplishments that we're proud of

Made DeFi Actually Accessible - Non-crypto friends understood EtherFi in 5 minutes
Production-Ready Code - Proper error handling, security, deployed worldwide
Seamless AI Integration - Feels like learning from a patient tutor
Interactive Learning - Calculator, quiz, and adaptive modes
Polished UI/UX - Professional design built in 3 days
Real Impact - Solves genuine education barriers in DeFi adoption


What we learned

Technical:

  • AI prompt engineering for educational content
  • Multi-API integration with error handling
  • Streamlit state management and deployment
  • Security best practices (environment variables)

Domain:

  • Deep dive into liquid staking mechanics
  • Understanding restaking through EigenLayer
  • Token economics (eETH, weETH)
  • DeFi risk assessment and yield calculations

Soft Skills:

  • User-centric design thinking
  • Scope management under time pressure
  • Systematic debugging and problem-solving

What's next for EtherFi AI Tutor

Short-Term (2 weeks):

  • Enhanced quiz with difficulty levels and leaderboards
  • Multi-protocol comparison calculator
  • Portfolio integration (connect wallet, track earnings)

Medium-Term (2 months):

  • Structured learning paths (Beginner → Advanced)
  • Community features (share explanations, discussion forum)
  • Voice interaction for accessibility
  • Multi-language support

Long-Term (6+ months):

  • Direct staking through the interface
  • Personalized AI that learns from user interactions
  • White-label version for other protocols
  • Gamified DeFi simulation

Vision:

  • Make Eli the go-to educational resource for all DeFi protocols

Final Thoughts

DeFi can democratize finance, but only if people understand it. Education is the bridge between innovation and adoption.

Eli represents a new paradigm: AI-powered, personalized, always-available education that meets learners where they are. This isn't just a tech demo it's a tool that could genuinely help thousands of people benefit from protocols like EtherFi.

Built with by Venkat Preetam Pulla for the ASU Claude Hackathon

Built With

  • anthropic
  • claude
  • etherfi
  • python
  • streamlit
Share this project:

Updates