Inspiration

Know anyone that has been affected by the recent tech lay offs?

Many people in the growing tech industry are anxious and threatened about job security, causing many to work multiple jobs. In the year of 2023, multi-job employees has increased by 5%, or in other words, 4.6 million people in the US.

However, adjusting to the lifestyle of increasing working hours and decreasing attention to the social life that you’re used to is something that is not easy to get used to.

Obviously, we can’t help increase hiring rates, but we did find a solution to helping others juggle multiple jobs and projects.

What it does

Parsely allows for processing and analysis of multiple documentation types. It can reorganize, analyze and create task lists just from your uploaded files. On top of that, Parsely allows for entrepreneurs to build their own customer representative chatbots for their own businesses, allowing for product recommendation, product quoting and booking. From document reorganization and analysis to customer chatbots, your time can now be spent on things you value more-- your friends, your family and most importantly, yourself.

In this tough time for all of us, we can still make the most of it, especially since we are more than just a garnish.

2.6.24 Update:

  1. Parsely supports 20+ file types, easily answering multiple questions over a wide diversity of file contents accurately. 
  2. Data is stored in two ways: (1) User session based, which uses the session state to store all data. (Finished) (2) User database based, which uses Cloud Vector DBs to help customers deploy their personal product recommendation and booking agency and personal smart data retrieval vault. (WIP)
  3. Looking for Cloud VDB's credit sponsorships: The credits will be used to help host the alpha test users' data for a feature called BYOB. A unique link will be generated using the user's cloud api key and collection name to allow the test user to deploy their shop sales consultant and booking agent using their product database and business document.
  4. People are likely going to want to have control over which database they want to use (ie Qdrant vs Vectara vs Astra vs so many others) so that they own their own contents. Parsely make professionals who don't know how to code to have a leverage over people who could code their own personal assistance over all of their data. - Plus, they get to save their money on hundred thousands/millions of dollars over their own tech teams for the less "important agenda" projects when it comes to data transformation and retrieval and augmentation.
  5. So far, our use cases have saved at least $12k to $50k per project, for 4 different projects. The cost was trivial in comparison, less than $200 cost throughout a three months scope. 

How we built it

Over the weekend, our team dedicated efforts to developing comprehensive API endpoints for our documentation analysis. We undertook a significant overhaul of our Streamlit skeleton, fortifying our backend infrastructure to ensure a more robust system with reduced overall runtime. With successful storage of all file types into the session state and nodes, our focus shifted to meticulously parsing document names and their associated metadata content from the Llama_Index Document Nodes, employing the effective 'ic(key, v.text)' approach. Following this, we delved into crafting the summary and achieved noteworthy progress in completing the initial page of the script.

Additionally, parallel to our technical advancements, we engaged in user-centric activities, including user research, sketch creation, user flow development, and the generation of quick, preliminary wireframes and prototypes. These endeavors aimed to offer a tangible representation of the anticipated user experience, providing valuable insights into the overall aesthetics and functionality of the platform at its foundational level. Throughout the weekend, our primary focus was on refining the platform's functionality and optimizing the backend to deliver an enhanced user experience.

Accomplishments that we're proud of

Being a small 2 person team, we have pushed through the weekend to get our features out and running. Even with the lack of a front end developer, we still managed to learn and most importantly have fun

What we learned

Gaining insights from various perspectives, including those of end users, and engaging in conversations about feedback and pain points proved invaluable for obtaining a comprehensive understanding of the user experience. Throughout the development process, maintaining humility in recognizing the diverse range of products in use became paramount. Acknowledging that the latest innovations only become truly beneficial when effectively implemented, we emphasized the significance of practicality and applicability.

A central focus during development was identifying approaches that yield maximum time-saving leverage in proportion to the invested time. Striving for efficiency, we prioritized solutions that offer the most significant time savings while remaining conscious of cost-effectiveness and speed in deployment. This approach ensured that our efforts were aligned with both optimal development practices and the practical needs of our users.

What's next for Parsely

On top of polishing the user experience and decreasing the run time, our first next action item is adding download and prompt copying options. This allows for user accessibility where generated data can be recorded in a different file or shared with others. We also want to create more organization features in the navigation bar, being mindful that multiple prompts and uploads will be generated. Instead of having a list of past prompts, was want to create prompt tags that can help the user easily find their needed uploads and documents. We also want to implement a general analysis report for uploaded files, allowing for the user to easily view well organize details and statistics.

In addition, we have been thinking about making it a web extension. Because this product is so malleable through different industries such as education, finance, management and more, we believe that accessibility is key. Through a web extension, our platform would easily get your job(s) done without having to switch tabs.

Built With

  • bentoml
  • figma
  • llama-index
  • llama-parse
  • ocr
  • python
  • rag
  • streamlit
Share this project:

Updates