0% found this document useful (0 votes)
8 views19 pages

AI Chatbot for Personalized Finance

Uploaded by

ayush.singh1
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
8 views19 pages

AI Chatbot for Personalized Finance

Uploaded by

ayush.singh1
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

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


PParticipants 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 - ₹299
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 - ₹249


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 Confidence Justification


Recommendation

Plant-Based
"Vegan snacks 95% "Perfect vegan option within
Protein Cookies
under ₹300" budget with high protein content
(₹299)
for

health-conscious snacking"
Handblock Print
"Light 92% "100% cotton kurta keeps you
Kurta (₹1499)
ethnic cool while maintaining
wear for traditional elegance for summer
summer" occasions"

Oats & Chia Seed "15g plant protein per serving


"Protein-rich 88%
Bars (₹349) makes this ideal for active
breakfast
lifestyles and morning nutrition"
options"

Oversized Cotton Tee "On-trend oversized fit in


"Trendy casual 85%
(₹799) premium cotton, perfect for
wear under
casual outings within your
₹1000" 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: $30,000

• 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

1. Working Demo Video (max 3 minutes):

2. ​ GitHub Repository with:

3. ​ PPT

4. ​ 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 deploye


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
Submission Form Link :
[Link]
K0E1p-RarOJbcT4WgZCf_SrdB7e2K0PegbMd0w/viewform
?usp=publish-editor

You might also like