Business Plan Bot

Securing a loan is an important step for a small business. The tool aims at reducing racial biases associated with loan approvals by making prerequisites less intimidating to the end user.

Comment

Access to capital for minorities has historically been an unequal playing field in the United States. Informed by my experience growing up and translating documents for my family members, I recognize the value that language translation of basic concepts can provide. My application is designed to help overcome language barriers and a lack of familiarity with business concepts, with the goal of improving access to loans from traditional lending institutions for non-English-speaking members of minority populations.

The application leverages AWS Bedrock to take user input and generate business plans, financial forecasts, and best practices for securing personal identifiable information (PII). AWS's generative AI service, using the Titan Text Embeddings V2 language model, simplifies the phrasing of complex information while allowing users to request clarifications as needed. The knowledge base for the model includes 2KB of best practices for drafting business plans from academic pdf and the series by Michael Bazzell on Open Source Intelligence. The front end offers a simple layout, with language options tailored to communities typically under served by traditional lending institutions.

The front end was built using Vanilla JavaScript, while the back end utilizes Python with the Flask framework and Boto3 to interact with AWS services. For reference and inspiration on connecting Bedrock to a front end, I forked the repo from mikehostetler/amplify.

While developing the application with my work account, one of the key challenges I faced was configuring IAM permissions for AWS Bedrock to allow proper connection to AWS services. I was able to generate AI output using AWS, but due to permission constraints from my supervisor, I was unable to connect it directly to the front end.

Despite these challenges, I'm proud of configuring the language model to produce curated outputs on the AWS Console and of expanding my knowledge of AWS Bedrock.

As a result of my participation in Shellhack, I’ve gained extensive knowledge of the IAM policies required to connect AWS Bedrock to Boto3- for which I am very thankful.

I also learned that while applications for generating business plans are fairly common, few cater to non-English-speaking minority populations. My future plans include enhancing the application to allow users to upload budgets and ledgers, which the LLM can then use to generate financial projections—a common requirement for lenders. I also plan to present this project to management at my current job to explore the potential of using AWS Bedrock to upskill workers in financial concepts.

Share this project:

Updates