Agiler.

Let your projects track themselves, while you focus on building.

Comment

Inspiration

Our team wanted to solve for the product management space, so we met with a solutions architecture director at a consulting firm to get more insight on how companies use existing product management tools. To our surprise, it wasn't the tool that was the problem, it was the fact that people were not using the tools readily available to them.

What it does

Agiler. It is a product management tool that automates the tracking and updating. The current way of doing things is: teams will have their standup meetings or weekly checkins with various stakeholders, then the product/project manager will update the status of tickets or create new tickets. That sounds like busy work to me. Instead, Agiler can do this for PMs using recordings of meetings or meeting notes.

How we built it

Front end: React.js with Tailwind CSS. Back end: Python with FastAPI, Google Gemini for multi-modal, audio inputs and processing, and Anthropic Claude for text inputs and processing. Finally, we used Google O-Auth for log-in and Supabase. The site is deployed on Vercel, and the FastAPI server is deployed with Docker and Google Cloud Run.

Challenges we ran into

We initially used Google Gemini for JSON formatting (specifically, creating valid IDs), but we had trouble getting consistent output after a lot of prompt engineering. We then tried Anthropic Claude where we discovered that Claude was more suited for generating structured outputs, but Google Gemini was still superior when it came to multimodality.

Accomplishments that we're proud of

In general, we took critical issues that professionals in the field raised, and we implemented a solution that was able to solve the majority of them. More specifically, we used LLMs to directly interact with our database in order to populate and update the user story dashboard. Moreover, we recreated a functional product management software while incorporating advanced AI features to automate the busy work to assist PMs.

What we learned

In general, we learned to use cutting-edge development technologies and frameworks, as well as product management software/tools and the metrics that PMs need to keep track of. During development, we also discovered that despite many models being all-purpose, some are better than others for certain tasks. For example, we noticed that Gemini was better for multi-modal applications such as audio processing, and Claude was better for generating large, parsable, structured outputs.

What's next for Agiler.

Our team will polish the site and open it up for testing to get feedback on what features the app should have. We want to make Agiler an app product teams would actually consider using as an alternative to big names like Jira.

Built With

Share this project:

Updates