This Capstone Project serves as the culmination of the 7-week Financial AI course, where participants have progressively learned the essential components for building intelligent, AI-driven financial applications. Throughout the program, participants have explored the following topics:
After completing these modules, participants are expected to integrate and apply all the knowledge and techniques they have acquired into a real-world, deployable application.
The purpose of the capstone project is to simulate how AI projects are executed in professional environments. Participants must translate theoretical understanding into a working product that demonstrates problem-solving, technical proficiency, and creativity.
This project reflects real-world industry needs—how financial institutions leverage AI to automate workflows, summarize complex data, assist in decision-making, and enhance customer interactions. Through this capstone, students not only show their technical mastery but also experience the end-to-end lifecycle of an AI solution: from problem framing and data handling to model integration and deployment.
Ultimately, this capstone is designed to help participants:
This section guides you in planning and documenting your capstone project effectively. Each part below corresponds directly to the components that must appear in your Project Report, ensuring your documentation is structured, comprehensive, and aligned with the final grading criteria.
Start your report by clearly defining the real-world problem or business need your application aims to address. This helps readers understand the relevance and impact of your project.
You can structure your explanation using the 5W + 1H framework to make it comprehensive:
What – What is your project about? What specific problem or use case are you trying to solve?
Example: “The application focuses on automating credit risk assessment using financial document summaries.”
Who – Who are the intended users or stakeholders of your application?
Example: “The target users are financial analysts and loan officers who need fast access to summarized client data.”
Why – Why did you choose this problem or dataset? Why is it important or relevant?
Example: “Manual credit assessment is time-consuming and prone to bias; automation can improve efficiency.”
When – When was the data collected, and is it still relevant?
Example: “The dataset includes financial transactions from 2021–2024 to capture post-pandemic trends.”
Where – Where will users interact with your app (dashboard, chatbot, etc.)?
Example: “The app will be deployed on Streamlit Cloud, providing web-based accessibility.”
How – How does your application solve the problem?
Example: “By using an LLM to summarize unstructured reports and visualize KPIs on a Streamlit dashboard.”
This part sets the context and purpose of your application and helps ensure that your project aligns with real business challenges.
Define what you want to achieve through your application in one or two clear sentences. The objective should directly address the problem mentioned above and indicate measurable outcomes.
Example Objectives:
- To create a chatbot that answers financial queries using Retrieval-Augmented Generation (RAG).
- To automate financial data summarization and visualization in a single interactive dashboard.
- To build an LLM-powered tool that classifies loan documents into predefined categories.
Your objectives will act as the foundation for later sections—especially when explaining your system design, data flow, and evaluation.
Explain how you utilize Large Language Models (LLMs) in your application. Detail the approach you take to optimize its effectiveness and performance. Some aspects to include:
Example: “We used OpenAI GPT-4 via LangChain’s ReAct agent architecture to process user queries and retrieve financial summaries from a local CSV dataset.”
This part demonstrates your understanding of LLM application design and orchestration.
Document how your data is handled throughout the app — from raw input to final output. You can describe this in narrative and/or flow diagram formats.
Include:
Example: “The application reads financial transactions from CSV → cleans missing values → computes monthly aggregates → generates visual insights → LLM summarizes the trends.”
This section helps evaluators understand the logic and reproducibility of your workflow.
Describe the design principles and interaction flow of your Streamlit app. Focus on clarity, usability, and layout logic.
Cover the following points:
Example: “We applied a clean layout with key financial KPIs displayed as metrics on top, followed by visual summaries and chatbot interaction at the bottom.”
Explain how you deployed your app and what configurations were required.
Include:
streamlit, langchain, openai,
etc.)Example: “The app is deployed on Streamlit Cloud using Python 3.11 with API keys stored in Streamlit secrets.”
This final section of your report should be reflective and forward-looking. Write about:
Example: “We learned how RAG improves LLM accuracy but increases latency. Future work includes optimizing retrieval time and adding new financial datasets.”
After completing the project design and documentation (as outlined in the Project Overview), participants must prepare and submit two main deliverables:
Both deliverables are equally important — the app demonstrates your ability to build and deploy a functional solution, while the documentation shows your understanding of the entire development process and design rationale.
Your deployed Streamlit application is the final working product of your capstone. It should demonstrate your ability to design, implement, and deploy a functional AI-driven solution that addresses a real use case.
Must be publicly accessible, hosted on a platform such as:
The app should run without major errors and include at least one core functionality related to the course, such as:
Interface should be clear, intuitive, and logically structured.
The app must integrate at least one LLM or AI feature meaningfully.
The documentation notebook (Project_Report.ipynb) acts
as your technical narrative. It must explain
why and how you built your application
— including design rationale, workflow details, and insights gained
during the process.
Your capstone project will be evaluated based on the following six components. Each category reflects both the technical and conceptual understanding expected from a professional-grade AI project.
| No. | Evaluation Criteria | Description | Score |
|---|---|---|---|
| 1 | Problem Definition & Relevance | Clearly explains the problem being solved and its relevance in the financial (or other) domain. The use case should be realistic, impactful, and well-justified. | 15 |
| 2 | Documentation & Workflow Explanation | Well-structured notebook that walks readers through each step, includes reasoning behind design choices, and provides reflection on results. | 20 |
| 3 | Use of LLMs or AI Tools | Effective and purposeful use of LLMs or AI components (prompting, API, chaining, retrieval, memory, etc.) integrated within the app. | 20 |
| 4 | Streamlit App Design & Functionality | The app works as intended, features a clean and intuitive UI, and fulfills the stated use case or business goal. | 20 |
| 5 | Creativity & Technical Execution | Innovative or well-thought-out approach, efficient code structure, good use of tools and libraries, and maintainable implementation. | 15 |
| 6 | Deployment Completeness | The app is successfully deployed, publicly accessible, and runs without major technical issues. | 10 |
| Total: 100 |
Deadline: Friday, 7 November 2025
Files to Submit:
Link to your deployed Streamlit app
GitHub repository link containing:
requirements.txtProject report notebook
(Project_Report.ipynb)
Naming Convention:
Capstone_<Name>_<ProjectTitle>Example:
Capstone_John_FinChatDashboard
Submission Platform: Google Classroom
💡 Tips for Achieving a High Score:
- Keep your app simple but polished. Focus on clarity, not complexity.
- Ensure your LLM or agent integration is meaningful and demonstrable.
- Double-check all links, file paths, and dependencies before submission.
- Document your reasoning — explain why you made each design choice.
- Include visuals, diagrams, or screenshots in your report for better readability.
⚠️ Common Pitfalls to Avoid:
- Missing or broken deployment links
- App errors due to missing API keys or libraries
- Incomplete documentation (especially missing system diagram or reflection)
- Overly technical explanations without connecting to business use case
This section provides a collection of ideas and exploration directions to help you design and expand your Capstone Project. You are encouraged to be creative, explore multiple tools, and integrate various components — as long as they align with your financial AI use case and maintain technical soundness.
You may choose from a wide range of project types, depending on your interests and creativity. Below are some examples of Financial AI App variations that can be developed using Streamlit and LLMs:
| Type | Description / Example Use Case |
|---|---|
| AI Financial Insight Dashboard | An interactive dashboard that uses LLMs to summarize market trends, company performance, or portfolio health. Example: “Summarize top stock movements and explain drivers of change.” |
| AI Financial Visualization Tool | A dynamic visualization app that translates financial data into insights, e.g., plotting returns, correlations, or volatility with AI-based commentary. |
| AI Prediction Dashboard | Uses a predictive model (traditional ML or LLM-based reasoning) to forecast financial metrics such as loan risk, stock trend, or credit default probability. |
| AI Summarization App | Summarizes financial reports, news articles, or company filings (PDFs, websites, or uploaded text). Useful for investors or analysts. |
| AI Chatbot for Financial Q&A | Conversational agent that answers user queries about market trends, stock comparisons, or company financials using retrieval or API-based data. |
| AI Document Chatbot | Enables users to upload financial documents (PDF, CSV, or text) and interactively query them for summaries or explanations. |
| AI Portfolio Advisor / Analyzer | Provides risk assessment or diversification advice based on user input and external market data. |
| AI Report Generator | Automatically produces a structured report (Markdown or PDF) summarizing financial performance, predictions, and insights. |
💡 You can combine more than one variation (e.g., a financial chatbot inside a dashboard, or a summarization feature within a visualization app) to create a more comprehensive solution.
Integrating external tools and APIs allows your application to become more powerful, dynamic, and data-driven. Here are several directions for integration:
| Tool Type | Example Use / Benefit |
|---|---|
| Web Search / News APIs | Retrieve real-time financial news or stock-related articles for analysis or summarization (e.g., Bing Search API, Serper.dev). |
| Financial APIs | Fetch live or historical financial data for analysis.
Popular sources include: • Sectors.app API – industry & company-level data • Yahoo Finance – stock prices, earnings, trends • CoinGecko API – cryptocurrency data |
| Database Integration | Store and retrieve chat history, analysis results, or user preferences using SQLite, Supabase, or Google Sheets. |
| Data Analysis Tools | Utilize Python-based analytical tools within the app,
such as pandas, NumPy, yfinance,
or matplotlib, integrated through LLM reasoning or
tool-chaining. |
| Agent & Tool Integration | Combine multiple AI tools (e.g., web search + data analysis + summarization) using a framework such as LangChain or LangGraph. |
🧠 Tip: Use tool chaining to simulate “agentic” behavior — for instance, an agent that retrieves financial data, runs a pandas analysis, and then explains the results conversationally.
When combining multiple components, aim for a cohesive and maintainable system. Below are several practical integration strategies to explore:
LLM + API Integration
LLM + DataFrame Analysis
LLM + RAG (Retrieval-Augmented Generation)
LLM + Streamlit Dashboard
LLM + SQL / Vector Database
⚙️ Focus on functional synergy — each component should add value, not just exist independently.
Reliable data sources are essential for building meaningful financial AI solutions. Below are suggested APIs and repositories you can explore:
| Source / Platform | Use Case |
|---|---|
| Sectors.app API | Company and industry insights |
| Yahoo Finance API (yfinance) | Stock and index data, price history |
| CoinGecko API | Cryptocurrency market data |
| Alpha Vantage | Real-time stock and forex data |
| Finnhub.io | Financial statements and sentiment analysis |
| Kaggle Datasets | Static datasets for testing or prototyping |
| Investing.com | Market summaries and commodities prices |
💡 Always check the data license and API call limits before integrating them into your project.
Your Streamlit dashboard can be as simple or as complex as needed. Consider including interactive and user-friendly features to make your solution engaging and functional.
| Feature | Purpose / Example Implementation |
|---|---|
| Data Upload (CSV, Excel, PDF) | Allow users to upload their own financial or transaction data for analysis. |
| Dynamic Filtering / Sorting | Interactive tables to explore company metrics or loan data. |
| Automated Summarization | Generate a short insight or summary for each uploaded dataset. |
| Visual Analytics | Charts for portfolio distribution, trend lines, correlations, etc. |
| Conversational Agent Panel | Chat interface for querying or interpreting dashboard insights. |
| Report Generation | Export dashboard results as PDF, DOCX, or Markdown report. |
| Session Memory / History | Save past user queries and responses for continued conversations. |
| Theme Customization | Add toggle between light/dark mode or custom UI design. |
🎯 A good dashboard should tell a story — combine visuals, analysis, and AI-driven narrative.
The Capstone Project marks the culmination of your journey in the Financial AI Course, where all the skills, frameworks, and concepts you’ve learned come together in a practical, real-world application.
This final phase not only tests your technical ability but also your creativity, problem-solving mindset, and understanding of how AI can generate value in the financial domain.
Before final submission, take time to reflect on your development process and the decisions you made throughout the project. These guiding questions can help structure your reflection in the documentation:
Problem Understanding
Design and Decision Process
Technical Challenges & Solutions
LLM Utilization
User Experience & Functionality
Deployment & Maintenance
Learning Outcome
🧭 Remember: a strong reflection is not about perfection, but about awareness — what worked, what didn’t, and how you grew as an AI practitioner.
By completing this capstone project, participants are expected to demonstrate the following competencies:
| Area | Learning Outcome |
|---|---|
| Problem Framing | Ability to identify real-world financial challenges and translate them into AI-based use cases. |
| AI System Design | Capable of designing an end-to-end application architecture integrating LLMs, APIs, and data pipelines. |
| Technical Implementation | Implement solutions using Python, LangChain/LangGraph, and Streamlit with clean and reproducible code. |
| Analytical & Critical Thinking | Apply reasoning to interpret financial data, derive insights, and evaluate outcomes. |
| Prompt Engineering & Tool Chaining | Utilize prompting, structured output, and agent-based reasoning effectively. |
| UI/UX Awareness | Design interfaces that balance clarity, functionality, and interactivity for non-technical users. |
| Deployment & Documentation | Deploy a functional public app with proper setup, clear documentation, and reproducibility. |
| Professional Readiness | Translate AI prototypes into practical business solutions suitable for real financial contexts. |
Your capstone is just the beginning. Once you complete it, you’re encouraged to expand your project or turn it into a professional portfolio piece. Here are several directions to continue:
🌱 Think of your capstone as a seed — it may start as a simple demo, but it can grow into a full-scale AI product with real business value.
This capstone project is not only a test of your technical ability but also a simulation of real-world AI solution development — where understanding the user, data, and system integration is just as important as writing the code.
As you finalize your submission, remember:
Maintaining academic integrity is a fundamental value of this course. Plagiarism, in any form, is strictly prohibited and will result in serious academic consequences.
Plagiarism refers to presenting someone else’s work, ideas, or materials as your own without proper acknowledgment. This includes, but is not limited to:
⚠️ Even partial copying — such as duplicating sections of a chart, analysis logic, or prompt structure — is considered plagiarism if not properly attributed.
To ensure fairness and originality, all students must adhere to the following principles:
Create Your Own Work All dashboards, analyses, and notebooks submitted must be developed and written by you. You may refer to examples provided in class but must build your own unique implementation.
Acknowledge All Sources If you use any reference material (datasets, code snippets, or ideas), always cite the source clearly in your report or notebook comments. Example:
Use AI Tools Responsibly You are allowed to use AI assistants (such as ChatGPT, Copilot, or Gemini) to support your learning, but not to submit their raw output as your own. You must critically review, modify, and integrate any AI-generated content before submission.
Avoid Code or Visualization Reuse from Peers Each participant’s submission must be distinct. Collaborative discussions are encouraged, but code sharing or duplication between students is not allowed.
Originality in Design & Content
Violations of academic integrity will lead to disciplinary action, including but not limited to:
All submissions will be subject to plagiarism and originality checks, including review of code history, notebook metadata, and dashboard inspection logs.
💡 Remember: the goal of this capstone is not to produce a “perfect” app, but to demonstrate your genuine understanding and ability to apply AI techniques in a real-world context.
Academic integrity reflects professional integrity. In the financial industry, trust and transparency are vital — the same principles apply to this capstone.
Your honest effort will not only help you pass this course but also strengthen your credibility as a future AI professional.
🧭 Be original. Be ethical. Build with integrity.