PRISM

Inspiration

When the theme of space economy was announced for this challenge, we were all confused as we had no experience and knowledge about space investments. Our newfound curiosities drove us to research reports from McKinsey and other outside sources and we were surprised to see that the overall space economy had grown from nothing to over $400 billion in just half a century. Our "lightbulb moment" was when we realized that we could just cross-reference the government space economy data with the 2020 COVID pandemic shock to discover how each industry reacted and which sectors actually survived the shock. We saw that airlines and tourist industries collapsed, while satellite communications, GPS systems, and space infrastructure were critical during the lockdown, but the Bureau of Economic Analysis data which we were given revealed much more nuanced resilience patterns. This inspired us to build an AI-powered investment advisor that combines local LLM technology with rigorous statistical analysis to democratize complex government data, helping investors make evidence-based decisions rather than following space industry hype.

What it does

The Space Economy Investment Advisor is an AI-powered platform for casual conversation with space economy data analyzed by the Bureau of Economic Analysis with the data set spanning the period 2012-2023 for real-time insight for investments. Combines local LLM and custom R statistical analysis techniques to offer an interactive chatbot that answers complex questions about space industry trends, growth patterns, and investment opportunities. Users can have natural conversations concerning particular sectors, ask for recent analyses of the latest data, or explore resilience metrics showing how different sectors of the space economy performed under the 2020 pandemic shock. Key functionalities of the system are the highest space sectors by composite score for investability, growth potential, and market resilience; forecast analysis with MAPE to check predictability; and extensive backtest results to validate investment strategy. The system runs 100% locally without any cloud dependency, safeguarding privacy and giving sophisticated technical financial analysis.

How we built it

We packaged the AI process analyses with an interactive web visualization. In its most basic incarnation, the system is a Streamlit chatbot that locally integrates an LLM (Ollama llama3.2:3b) for conversational investment advice, while 12 years of Bureau of Economic Analysis data in space economy are processed with custom R scripts. Weights for MAPE analysis of investment potential are computed at the backend using composite investment scores from growth potential, market resilience, and predictability criteria. The interactive HTML dashboard, designed for the client, provides dynamic user interactivity through Plotly.js with visualization types from heatmap performance, growth-resilience scatter plots, and historical trends. The system supports synchronous Python API handling, R script launching using subprocess, and real-time data processing. While both platforms share the same analytical engine, each appeals to different needs: The chatbot supplies personalized, conversational insights, while the website supports the need for heavy visual aids. The local deployment of the architecture prevents data from traversing cloud platforms, thereby guaranteeing complete data privacy while providing a very detailed financial analysis via user-friendly interfaces for the investor and researcher.

Challenges we ran into

Coordinating all these different technologies (Python Streamlit, R analytics, local LLM, and HTML dashboards) to arrive at a working system in which data flows uninterrupted through all the components. Managing local deployment hell-land, especially getting Ollama LLM to play well with your Streamlit front end, connection timeout issues, or model availability questions. Processing and accounting for 12 years of Bureau of Economic Analysis space economy data to derive some solid investment insights that we can then turn into conversational AI responses as well as interactive visualizations that actually mean something to the user.

Accomplishments that we're proud of

We have managed to make this complete dual-platform solution that brings together conversational AI and interactive data visualization-an interface integrating a local LLM (Ollama) with real-time BEA space economy data analysis through a custom R analytics engine. Our system provides an intelligent chatbot through Streamlit for natural language investment queries and a fancy HTML dashboard with Plotly.js visualizations for a fully-fledged advisory platform that validates 12 years of government data locally without ever making a call to the cloud. This project showcases great technical finesse by bringing together Python, R, JavaScript, and AI into one high-impact investment advisory service, made available through professional space-themed interfaces that help decode complex economic data for users.

What we learned

One of the most important lessons that came out of this was how quite difficult real-world data integration might be. For instance, data sets provided by governments such as BEA space economy data are time-consuming to preprocess, parse, error-check, and finally extract insights from. Hence we learned the importance of robust data-pipeline design. The project proved the existence of strong hybrid systems when multiple AI technologies were brought together (local LLMs-top statistical analysis-web interfaces) but also highlighted the imperative need for fallback mechanisms to kick in gracefully whenever such component-level failures occur-consistency in functionality from the user's perspective should never become conditional on technical issues. Even more striking, we found a very wide gulf between analytics power and what end-users can practically access-building the deep R analytics is only half the battle; the real value is in finally creating an interface that takes quite complex economic insights and turns them into everyday actions for the users with natural language interaction. We realized that the success of data science projects demands both technical penetration and design of user experience for true value delivery.

What's next for PRISM

• Expand our resilience-testing methodology to other emerging industries • Develop sector-specific investment timing models • Create an investor education platforms