Skip to main content

Hackathon Rabbitt AI

1 - Personalized Banking & Insurance AI Advisor​

Theme​

Build an intelligent AI chatbot advisor that provides personalized banking and insurance recommendations for Indian consumers. The system must understand user risk profiles, analyze financial needs, and recommend appropriate BFSI products with clear rationale - all within the Indian financial landscape.

Core Challenge: Context Engineering​

Participants must create an AI agent that:

  • Understands individual financial profiles and risk tolerance
  • Analyzes personal financial situations in the Indian context
  • Recommends tailored BFSI products with clear reasoning
  • Explains product benefits specific to user's circumstances

Target Audience​

  • Primary User : Average Indian adult consumer seeking personalized financial guidance
  • Age Range : 25-45 years
  • Income Levels : ₹3L - ₹15L annually
  • Financial Goals : Savings, insurance, investments, retirement planning
  • Language : English (primary requirement)

Product Categories to Cover​

1. Insurance Products​

  • Life Insurance : Term, endowment, ULIP policies
  • Health Insurance : Individual, family floater, critical illness
  • General Insurance : Motor, home, travel insurance

2. Investment Products​

  • Mutual Funds : Equity, debt, hybrid funds
  • Fixed Deposits : Bank FDs, corporate FDs
  • PPF/EPF : Tax-saving instruments
  • NPS : Retirement planning

3. Banking Products​

  • Savings Accounts : Regular, salary, zero-balance
  • Credit Cards : Cashback, rewards, travel cards
  • Personal Loans : Instant loans, gold loans
  • Home Loans : Interest rates, eligibility

Conversational Flow​

Advisor: "Hello! I'm your personal financial advisor. Let me understand your profile first."

User: "I'm 28, earning ₹60,000/month, want to buy term insurance"

Advisor: "Perfect! For a 28-year-old with ₹60K income, I recommend ₹75L-₹1Cr term coverage.

Here's why:

  • Your annual income: ₹7.2L
  • Recommended coverage: 10-15x income
  • Premium: ~₹8,000/year
  • Top options: HDFC Click 2 Protect, LIC Tech Term

Sample Implementation Requirements​

Core Architecture Expectations​

Semantic Search Implementation:

  • Use embeddings to find relevant financial advice from knowledge base
  • Implement similarity matching with confidence scoring
  • Enhance queries with user profile context for better results

Personalized Response Generation:

  • Generate contextual advice using LLM capabilities
  • Integrate user profile data into response formulation
  • Maintain professional financial advisor tone and expertise
  • Include specific product recommendations with reasoning

Expected System Flow:

  1. User profile analysis and risk assessment
  2. Query understanding and context enhancement
  3. Knowledge base search with personalization
  4. Response generation with Indian financial context
  5. Confidence scoring and source attribution

Key Deliverables​

  1. Working demo Chatbot Video (max 3min):
  2. Github repo
  3. Hosted Website url(if done)
  4. PPT

2 - Personalized E-Commerce Recommender Bot​

Theme​

Build an AI-powered recommendation assistant for Indian D2C food and fashion brands that delivers hyper-personalized product suggestions based on user context and preferences. The system must understand natural language queries, extract user preferences, and provide contextually relevant product recommendations with clear reasoning.

Core Challenge: Context Engineering​

  1. Understands User Context : Extracts key preferences (dietary needs, style preferences, budget) from natural language queries
  2. Personalized Recommendations : Matches products to user profiles using hybrid filtering approaches
  3. Explains Recommendations : Provides clear, brand-aligned reasoning for product suggestions
  4. Adapts to Indian Market : Considers local pricing (₹), seasonal trends, and cultural preferences

Conversational Interface​

Shopping Assistant: "Hi! I'm your personal shopping assistant. What are you looking for today?" User: "I need vegan snacks under ₹300" Shopping Assistant: "Perfect! Here are my top recommendations for you: Plant-Based Protein Cookies - ₹ 95% Match Why it's perfect: These gluten-free cookies fit your vegan diet and budget perfectly. With 12g plant protein per serving, they're both healthy and delicious! Quinoa Energy Bites - ₹ 87% Match Why it's perfect: Vegan superfood snacks under your budget with natural sweetness

from dates and coconut."

Sample Queries & Expected Outputs​

User Query Top Recommendation Confidence Justification "Vegan snacks under ₹300" Plant-Based Protein Cookies (₹299) 95% "Perfect vegan option within budget with high protein content for health-conscious snacking" "Light ethnic wear for summer" Handblock Print Kurta (₹1499) 92% "100% cotton kurta keeps you cool while maintaining traditional elegance for summer occasions" "Protein-rich breakfast options" Oats & Chia Seed Bars (₹349) 88% "15g plant protein per serving makes this ideal for active lifestyles and morning nutrition" "Trendy casual wear under ₹1000" Oversized Cotton Tee (₹799) 85% "On-trend oversized fit in premium cotton, perfect for casual outings within your budget"

Key Deliverables

  1. Working Demo Video (max 3 minutes):
  2. GitHub Repository with:
  3. PPT
  4. Hosted Website url(if done)

3 - Automated Data Filler Agent​

Theme Build an intelligent AI agent that automatically processes sales meeting summaries and populates CRM systems with structured data. The system must extract key information from unstructured text, understand sales context, and create accurate CRM entries - all within the B2B sales workflow.

Core Challenge: Context Engineering This is a Context Engineering Hackathon where participants must create an AI agent that:

  • Understands unstructured sales meeting summaries and conversations
  • Extracts key business entities, relationships, and deal information
  • Structures data into CRM-compatible formats with proper categorization
  • Integrates seamlessly with CRM platforms while maintaining data accuracy

Target Audience Primary User: B2B Sales Representatives and Sales Managers

  • Company Size: SMEs to Enterprise (50-5000 employees)
  • Industries: SaaS, Manufacturing, Consulting, Financial Services
  • Current Pain: Spending 2-3 hours daily on manual CRM data entry
  • CRM Platform: HubSpot, Salesforce, or Zoho

CRM Data Categories to Extract 1. Contact Information

  • Personal Details: Name, designation, email, phone number

  • Professional Info: Decision-making authority, department, seniority

  • Communication: Preferred contact method, availability, timezone 2. Company Information

  • Basic Details: Company name, industry, size, website

  • Business Context: Current challenges, pain points, technology stack

  • Market Position: Competitors, budget range, growth stage 3. Deal Information

  • Opportunity Details: Deal value, timeline, decision criteria

  • Sales Stage: Current stage, next steps, proposal requirements

  • Competitive Landscape: Alternatives being evaluated, key differentiators

Conversational Flow Agent: "Please paste your meeting summary, and I'll extract the CRM data automatically."

User: "Had a call with Sarah Johnson, Marketing Director at GrowthTech. They need marketing automation for 50-person team. Budget: $30K annually. Evaluating HubSpot vs our solution. Next: Demo scheduled for Friday."

Agent: "Perfect! I've extracted and structured your CRM data

: Contact Created:

  • Name: Sarah Johnson
  • Title: Marketing Director
  • Company: GrowthTech
  • Stage: Qualified Lead

Company Created:

  • GrowthTech (Marketing Industry)
  • Team Size: 50 employees • Budget: $30,000 annually

Deal Created:

  • Marketing Automation Implementation
  • Value: $30,
  • Stage: Demo Scheduled
  • Competitor: HubSpot
  • Next Action: Demo on Friday"

Sample Implementation Requirements Core Architecture Expectations Text Processing Implementation:

  • Use NLP to extract entities (names, companies, amounts, dates)

  • Implement relationship mapping between contacts, companies, and deals

  • Handle various meeting summary formats and writing styles

  • Validate extracted data against CRM field requirements

CRM Integration Implementation:

  • Generate API calls to create contacts, companies, and deals
  • Handle duplicate detection and data merging
  • Implement error handling for failed API calls
  • Maintain data consistency across CRM entities

Expected System Flow:

  1. Meeting summary input and preprocessing
  2. Entity extraction and relationship mapping
  3. Data validation and CRM field mapping
  4. API integration and CRM data creation
  5. Confirmation and error handling with user feedback

Key Deliverables

  1. Working Demo Video (max 3 minutes):
  2. GitHub Repository with:
  3. PPT
  4. Website link (If deployed)

4 -Personalized Learning Recommender for Coding Students​

Theme​

Build an intelligent AI-powered learning coach that analyzes coding solutions and provides personalized learning recommendations for college students practicing on platforms like LeetCode. The system must understand individual coding patterns, identify knowledge gaps, and create tailored learning paths with relevant resources - all focused on improving programming skills and interview readiness.

Core Challenge: Adaptive Learning Intelligence​

Participants must create an AI agent that:

  • Analyzes code quality, patterns, and problem-solving approaches
  • Identifies specific knowledge gaps and areas for improvement
  • Recommends personalized learning resources and study plans
  • Tracks progress and adapts recommendations based on learning patterns
  • Generates step-by-step improvement roadmaps for coding skills

Target Audience​

Primary User : College students preparing for coding interviews and improving programming skills

  • Education Level : Undergraduate/Graduate Computer Science students/ ANYONE who’s curious
  • Experience Range : Beginner to Intermediate programmers
  • Goals : FAANG interview preparation, competitive programming, skill development
  • Platform Focus : LeetCode, HackerRank, CodeChef problem solving

Problem Categories to Analyze​

1. Data Structures & Algorithms​
  • Array/String Manipulation : Two pointers, sliding window, prefix sums
  • Linked Lists : Traversal, reversal, cycle detection
  • Trees & Graphs : DFS, BFS, traversal patterns
  • Dynamic Programming : Memoization, tabulation, optimization
  • Sorting & Searching : Binary search, merge sort, quicksort
2. Coding Patterns​
  • Two Pointers : Fast/slow pointer, left/right pointer techniques
  • Sliding Window : Fixed/variable window problems
  • Backtracking : Permutations, combinations, constraint satisfaction
  • Greedy Algorithms : Optimization problems, scheduling
  • Divide & Conquer : Recursive problem solving

Sample Conversational Flow​

AI Coach : "Hello! I'm your personalized coding mentor. Let me analyze your recent LeetCode submissions to create a tailored learning plan." User : [Uploads code solution for "Two Sum" problem] AI Coach : "Great! I analyzed your Two Sum solution. Here's what I found: Strengths: • Clean variable naming and good code structure • Correct algorithm implementation with O(n) time complexity • Proper handling of edge cases Areas for Improvement: • Hash Map Patterns : You solved this correctly, but similar problems like 'Three Sum' require advanced hash map techniques • Space-Time Tradeoffs : Consider when to optimize for space vs time

Personalized Learning Plan:

  1. Next Challenge : Try 'Three Sum' (LeetCode #15) - builds on your hash map knowledge
  2. Concept Deep Dive : Hash map collision handling and optimization
  3. Resources : ○ YouTube: 'Advanced Hash Map Techniques' by NeetCode ○ Article: 'Hash Table Implementation Details' ○ Practice: 5 similar hash map problems Weekly Goal : Master hash map patterns (estimated 3-4 hours)" Key Deliverables
  4. Working Demo Video (max 3 minutes):
  5. GitHub Repository with:
  6. PPT
  7. Website link (If deployed)

5 - Personalized Learning Recommender for Math & English (Grades 1-5)​

Theme​

Build an intelligent AI-powered learning coach that provides personalized learning recommendations for elementary school students (Grades 1-5) in Math and English, specifically designed for CBSE/NCERT curriculum. The system must assess student knowledge levels, identify learning gaps, and create adaptive daily/weekly learning plans to improve foundational skills - with special focus on Rural India's educational needs.

Core Challenge: Adaptive Elementary Education​

Intelligence​

Participants must create an AI agent that:

  • Assesses student knowledge levels through adaptive questioning
  • Identifies specific learning gaps in Math and English fundamentals
  • Generates personalized daily/weekly learning challenges and activities
  • Tracks progress and adapts difficulty levels based on performance
  • Provides engaging, age-appropriate learning experiences for young learners
  • Advanced Operations : Large Numbers, Factors, H.C.F And L.C.M., Tests Of Divisibility
  • Decimals & Fractions : Fractions, Decimals, Averages, Percentages
  • Practical Math : Simple Interest, Profit and Loss, Time, Unitary Method
  • Geometry : Mensuration, geometry, weights and masses, symmetry
  • Measurement : Metric Measures, Bills, Temperature

Sample Learning Journey​

Initial Assessment Flow: AI Tutor : "Hi! I'm your learning buddy! Let's start with some fun questions to see what you already know!" Math Assessment (Grade 3 Example):

  • "Can you solve: 25 + 17 = ?"
  • "Which is bigger: 1/2 or 1/4?"
  • "How many sides does a triangle have?" Personalized Learning Plan: AI Tutor : "Great job! I can see you're good at addition but let's work on fractions together. Here's your learning plan for this week: Monday - Wednesday : Fraction Fun • Activity: Pizza slice fractions game • Challenge: Compare 1/2 and 1/4 using pictures • Practice: 5 fraction comparison problems Thursday - Friday : English Reading • Story: 'The Brave Little Mouse' (150 words) • Activity: Find 10 new words and their meanings • Writing: Write 3 sentences about your favorite animal Weekend Challenge : Math + English Mix • Create a story problem using fractions • Example: 'Ram ate 1/4 of a pizza. How much pizza is left?'"
  • Offline Capabilities : Downloadable content for areas with limited connectivity
  • Regional Language Support : Hindi translations and voice support
  • Cultural Relevance : Examples and contexts familiar to rural students Key Deliverables
  1. Working Demo Video (max 3 minutes):
  2. GitHub Repository with:
  3. PPT
  4. Website link (If deployed)

6 -Prompt Analyser​

Theme Build an intelligent AI agent that monitors brand visibility across major AI platforms and search engines. The system must generate relevant prompts, analyze brand mentions in AI responses, track competitor performance, and provide actionable visibility insights all within the evolving AI-powered search landscape.

Core Challenge: Context Engineering This is a Context Engineering Hackathon where participants must create an AI agent that:

  • Generates industry-specific prompts that users commonly search
  • Analyzes brand mentions and rankings across multiple AI platforms
  • Compares brand visibility against key competitors
  • Scores overall brand presence with actionable insights and recommendations

Target Audience Primary User : Brand Managers and Digital Marketing Teams

  • Company Size : Startups to Enterprise (50-10,000+ employees)
  • Industries : FinTech, E-commerce, SaaS, Healthcare, EdTech
  • Marketing Budget : ₹5L - ₹50L annually on digital marketing
  • Pain Point : Unknown brand visibility in AI-generated search results

AI Platform Categories to Monitor

1. Conversational AI Platforms

  • ChatGPT : OpenAI's flagship conversational AI

  • Google Gemini : Google's advanced AI assistant

  • Perplexity AI : AI-powered search and answer engine

2. Search-Enhanced AI

  • Microsoft Copilot : Bing-integrated AI assistant
  • Meta AI : Facebook's conversational AI platform

Conversational Flow

Analyzer: "Welcome! Let's analyze your brand's AI visibility. What's your brand name and industry?"

User: "PolicyBazaar, insurance comparison platform"

Analyzer: "Perfect! I'll analyze PolicyBazaar's visibility across ChatGPT, Gemini, and Perplexity using insurance-related prompts.

Generated Prompts (20 samples):

  • Best insurance comparison platforms in India
  • Where to buy car insurance online
  • Compare health insurance policies

Visibility Analysis :

  • ChatGPT : 7/10 prompts mention PolicyBazaar (Rank #2 avg)
  • Gemini : 5/10 prompts mention PolicyBazaar (Rank #3 avg)

Competitor Analysis : • Bajaj Allianz: 6/10 average mentions • HDFC ERGO: 4/10 average mentions • ICICI Lombard: 5/10 average mentions

Overall Visibility Score: 67/100 "

Sample Implementation Requirements Core Architecture Expectations Automated Prompt Generation:

  • Generate 20-30 industry-specific prompts based on brand category
  • Include variations: direct questions, comparison queries, recommendation requests
  • Simulate real user search patterns and intent
  • Cover different stages of customer journey (awareness, consideration, purchase)

Multi-Platform Analysis Implementation:

  • Integrate with 3+ AI platforms via APIs or web scraping
  • Extract brand mentions, rankings, and context from responses
  • Identify competitor mentions and relative positioning
  • Track sentiment and recommendation strength

Expected System Flow:

  1. Brand profile setup and competitor identification
  2. Automated prompt generation based on industry keywords
  3. Multi-platform query execution and response collection
  4. Brand mention extraction and ranking analysis
  5. Competitor comparison and visibility scoring with actionable insights

Key Deliverables

  1. Working Demo Video (max 3 minutes)
  2. GitHub Repository
  3. **Hosted Website url(if done)
  4. PPT

7 - AI Video Generation Challenge: Visionary Leaders Edition​

Problem Statement:​

Welcome to the AI Video Generation Challenge! In this creative contest, participants will use AI tools to generate a short video (45–90 seconds) that tells a coherent, visually consistent story featuring consistent character avatars , a stable background , and a unified tone throughout the entire video.

Theme: "AI Visionaries Unite: Leaders Speak the Future"​

Your challenge is to create a short AI-generated video featuring 5-10 AI company leaders discussing how their companies are revolutionizing the world through artificial intelligence. Using face avatar technology, bring these industry pioneers together in a compelling narrative about the future of AI.

The Creative Vision:​

Imagine the world's most influential AI leaders coming together to share their company visions and discuss how AI is transforming our world. Whether it's a panel discussion, a futuristic

summit, a collaborative presentation, or an innovative format you dream up — the creative possibilities are endless!

  • Rabbitt AI - AI-powered consumer devices
  • OpenAI - GPT and advanced AI systems
  • Anthropic - Constitutional AI and safety
What Your Video Should Explore:​
  • Company Visions : How is each leader's company changing the world?
  • AI Innovation : What breakthrough technologies are they developing?
  • Future Impact : How will their AI solutions transform daily life, work, or society?
  • Collaboration vs Competition : How do these leaders view the AI landscape?
  • Ethical Considerations : What responsibilities do they feel as AI pioneers?
  • Advice for the Future : What wisdom do they share about AI development?
Creative Format Ideas:​
  • Virtual AI Summit : Leaders presenting at a futuristic conference
  • Round Table Discussion : Intimate conversation between visionaries
  • Time Capsule Message : Leaders speaking to future generations
  • Collaborative Presentation : Building on each other's ideas
  • Debate Format : Friendly disagreement on AI's future direction
  • Documentary Style : Interviews with multiple perspectives
  • Futuristic News Panel : Leaders as expert commentators
  • Or Create Your Own Unique Format!

What We’re Evaluating:​

Criteria Description Consistency Is the avatar and background visually stable across the entire video? Creativity How original and fun is the idea? Narrative & Clarity Does the video tell a complete, coherent story in 45–90 sec?

Tone Consistency Is the tone (humor, suspense, emotion) consistent from start to finish? Technical Execution Smooth transitions, audio-visual sync, clean output Voice & Music (Optional) Are voiceovers expressive and relevant? Is background music used well?

Hints​

1 - Personalized Banking & Insurance AI Advisor​

Technical Stack​

Component Tool Rationale

Embeddings Qwen/mistral/llama(search the model on huggingface)

Fast API calls, no local setup

Similarity Search Cosine similarity (NumPy) Simple, no database required

LLM Groq or Gemini Reliable, fast responses

UI Framework Streamlit Single-file deployment

Data Storage JSON files/VectorDB Quick setup/Database setup

Context Management Session state Built-in Streamlit feature

2 - Personalized E-Commerce Recommender Bot​

Technical Stack​

Component Tool Rationale

Embeddings (^) sentence-transformers/all -MiniLM-L6-v Local model, zero latency, good performance Similarity Search Scikit-learn cosine_similarity No database setup, fast computation LLM Groq (Llama 3 70B) 500+ tokens/sec speed, excellent reasoning UI Framework Streamlit Single-script deployment, built-in chat Data Storage Static JSON No preprocessing, immediate availability Context Management Session state Built-in Streamlit session handling

3 - Automated Data Filler Agent​

Technical Stack​

Component Tool Rationale

NLP Processing GeminiI/Groq API Fast entity extraction, no local setup

Entity Recognition spaCy + Custom Models

Accurate business entity detection

CRM Integration HubSpot/Salesforce API Direct data creation and updates

UI Framework Streamlit Single-file deployment

Data Validation Pydantic Models Ensure CRM data quality

Workflow Engine Python/FastAPI Handle processing pipeline

4 -Personalized Learning Recommender for Coding Students​

Technical Stack Requirements​

Component Recommended Tools Rationale Code Analysis Python AST, Tree-sitter Parse and analyze code structure Embeddings Gemini/GroqI/Hugging Face models

Code similarity and pattern recognition

Vector Search FAISS, Pinecone, or Weaviate Efficient similarity matching LLM Integration Groq, or Gemini Natural language explanations

UI Framework Streamlit, or React Interactive learning dashboard Data Storage JSON/SQLite for prototypes User progress and recommendations Code Execution Docker containers/sandboxes Safe code testing environment

5 - Personalized Learning Recommender for Math & English (Grades 1-5)​

Technical Stack Requirements​

Component Recommended Tools

Rationale

Assessment Engine Python/JavaScript Adaptive questioning algorithms NLP for English spaCy, NLTK Text analysis and grammar checking Math Problem Generator SymPy, NumPy Dynamic math problem creation Learning Analytics Pandas, Matplotlib Progress tracking and visualization

UI Framework Streamlit, React Child-friendly interface design Voice Integration Web Speech API Audio support for young learners Database SQLite/PostgreSQL Student progress storage Curriculum Mapping JSON/YAML NCERT syllabus alignment

6 - Prompt Analyser​

Technical Stack​

Component Tool Rationale AI Platform APIs OpenAI, Gemini, Perplexity Direct access to AI responses

Prompt Generation GPT-4/Claude API Automated relevant prompt creation Data Processing Pandas + NumPy Analyze mention frequency and rankings Visualization Plotly/Streamlit Charts Interactive dashboards and scoring Web Scraping Selenium/BeautifulSoup Extract AI platform responses Competitor Analysis

Custom NLP Models Identify and rank competitor mentions

7 - AI Video Generation Challenge: Visionary Leaders Edition​

Tools:​

Participants are encouraged to use any of the following AI tools to create their videos:

Video Generation Platforms:

  • Suno AI – for generating AI music (optional background soundtracks)
  • Pika Labs – video generation with prompts (great for avatars and scenes)
  • ImageFX by Google – for generating specific image frames for scenes
  • Runway ML (Gen-2) – highly advanced for creating scene-consistent AI videos
  • Hailuo AI – for avatar-driven storytelling
  • ElevenLabs – for generating high-quality voiceovers with personality

Optional Add-ons:​

  • D-ID or Synthesia – for animated avatars with lip-sync
  • HeyGen – for avatar-based video narration
  • CapCut / Descript / Adobe Express – for final editing and scene stitching