RegLens– The market’s lens on regulation

Inspiration

We were fascinated by the capabilities of Generative AI and could not wait to apply it in the world of finance. We acknowledge the immense impact of regulation of on the stock market. Yet, it can be overwhelming to be kept up with the latest news and it is difficult to read hundreds of pages of a bill without a law degree. Financial markets are being reshaped by an unprecedented wave of regulatory changes — from AI governance to climate policy and trade sanctions. By leveraging Large Language Models (LLMs), we can lift crucial details that would impact industries and stock prices and model our investing strategy. Our team asked: What if we could turn this complexity into clarity — automatically?

What it does

By simply passing in a law document, an LLM would perform a deep analysis, highlighting every important details and affected tickers of the S&P500. It lists every affected stock and gives a risk rating depending on its impact. The user can then make an informed decision about their investments. RegLens :

Reads laws and policies (PDF, HTML, XML, or plain text) from global jurisdictions.
Extracts key elements — affected sectors, companies, and enforcement dates — using generative AI and NLP.
Cross-references this information with S&P 500 company filings (10-K, 10-Q) and market data.
Simulates portfolio exposure by assigning an impact score per company and sector.
Visualizes actionable insights through an interactive web dashboard for decision-makers.

How we built it

By leveraging tools provided by AWS, we were able to create and deploy easy-to-use web application that simply prompts the user to drop a law/bill for analysis. For deployment, we use AWS Amplify, a website hosting platform, and Amazon Aurora DSQL, a database system to store the processed data to be shown to the user. To process our documents, we parse them through javascript parsers such as Cheerio. The data is then sent to our AI model, Deepseek in this case, to analyze that data. Our AI model is hosted on AWS bedrock and the open source model itself is found on Hugging Face. We also use Sagemaker to call the endpoint of that model. Once the data is analyzed by our model, we send it towards our database via SQL query. That database is then listed on the website.

Challenges we ran into

We found that AWS services are incredibly useful for our usecases. However, the learning curve for these tools was quite steep. We were able to overcome these challenges by getting our hands dirty and testing out these tools ourselves.

Accomplishments that we're proud of

We were proud of how our pipeline has come together. Teammates have worked on separate parts such as the frontend, document processing, AI model deployment, and database system, and we were satisfied at how it worked together as a uniform system at the end.

What we learned

We have learned about the vast ecosystem of AWS tools. It is an all-in-one platform that provides any web service that one can ask for. We have learned plenty yet we still have a lot more to learn after this hackathon.

What's next for RegLens– The market’s lens on regulation

We would want more interactivity with the user! Adding a chat functionality to reflect upon the impacts of a certain bill would deepen the user's understanding of its impact. In addition, a search system would simplify the user's ability to know about bills that are lesser known.